Overlooked weak structural connections support human cognition under nonlinear connectome scaling

arXiv:2505.24125 · q-bio.NC · Submitted 2025-05-30 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Genomics Radio. Generated commentary on the latest computational biology and genomics papers.

Ines: We're looking at the paper: "Overlooked weak structural connections support human cognition under nonlinear connectome scaling".

Title and authors: Ines: We're diving into the title and authors of "Overlooked weak structural connections support human cognition under nonlinear connectome scaling." It sets up the whole argument we just discussed about those weak links mattering.

Marcus: Yeah, the title is pretty direct. It tells you that these connections aren't overlooked; they are important, but they need a specific way of being interpreted—the nonlinear scaling.

Yuki: From my side, it sounds like this paper is trying to fix an old problem where we assumed everything in the connectome was equally important, which we know isn't true when you look at the data.

Ines: Precisely. It’s about moving from a simple binary graph idea to something much more complex, where the functional influence of those weak links isn't just their raw weight.

Marcus: The authors are using this nonlinear scaling parameter beta to see if we can capture that nuance, and they show that when beta is not one point zero, these weak connections actually provide better predictions for cognitive abilities.

Yuki: I wonder how this relates to the broader picture of brain development or species differences; does this scaling factor vary between different populations?

Ines: That's a big question. If the optimal beta value changes depending on what you are predicting, it suggests that different cognitive functions rely on different organizational principles in the brain structure.

Marcus: The authors found specific optimal values for general cognitive ability, around zero point seven zero three, and for memory, around zero point six zero two. These numbers are concrete results from their connectome prediction model.

Yuki: So the sample size they used has to be large enough to support these kinds of specific scaling parameters being robust across different groups.

Ines: And they did test this across a human connectome database, which gives them a massive dataset, which is crucial for seeing if these findings are generalizable beyond just one study.

Marcus: It's good they used such a large cohort, because the statistics on the predictive associations were what allowed them to make claims about general cognitive ability versus more specialized things like crystallized intelligence.

Yuki: I’m glad they specified that memory showed a stronger effect from this scaling than crystallized intelligence did, which supports the idea that memory might be more sensitive to these weak connections in this model.

Ines: That points towards a functional heterogeneity among those weak connections, meaning not all weak links are equally important for all cognitive tasks; some are more influential than others.

Marcus: So the paper's main contribution is showing that you need this scaling mechanism to unlock the predictive power hidden in those weaker structural connections. This sets up the next part where we look at how they actually found these findings.

The paper's summary: Ines: Okay, moving into the actual summary of "Overlooked weak structural connections support human cognition under nonlinear connectome scaling." They are explaining the mechanism behind why this matters for cognition.

Marcus: They argue that the brain isn't just a sparse graph where only strong connections count; it’s heavy-tailed, meaning there are many more weak connections than we usually think.

Yuki: So, they're saying these weak connections are an inherent part of the brain architecture, not just errors in the mapping process. They exist across several orders of magnitude in weight.

Ines: That’s it. And their summary is that under a linear weighting assumption—where beta equals one point zero—these weak connections have little effect on graph theoretical properties or neural signal propagation ratios.

Marcus: They say removing the weakest fifty percent of connections had negligible effects on the neural signal propagation ratio in a rat-brain dynamic model, which was a key piece of evidence against the idea that all weak connections are functionally irrelevant.

Yuki: So they're saying these weak links aren't just noise to be discarded; they are naturally pervasive features of biological networks, spanning several orders of magnitude in importance.

Ines: And their summary concludes that under the nonlinear scaling framework—when beta is less than one point zero, these weak connections can make disproportionately large contributions to brain functional organization and cognition despite their small absolute weights.

Marcus: That’s the big takeaway for us: weak connections are functionally relevant if you use the right mathematical interpretation of their weights. It’s not just about the raw tractography value.

Yuki: So, they are establishing a framework for reliably characterizing structural brain connectivity that accounts for this continuum of weights instead of just looking for a single cutoff point.

Ines: They also found that weak connections support both large-scale integration and fine-scale segregation simultaneously by expanding the modal repertoire of the brain network hierarchy.

Marcus: That means they’re not just making one kind of link, but they's supporting different levels of organization at once, which is a complex idea to grasp.

Yuki: I like that complexity because it suggests the brain is doing more than just one thing; it’s managing these competing organizational needs simultaneously.

Ines: And they pinpointed a specific subtype—those linking visual and motor systems with limbic regions and having negative gene co-expression—as having a disproportionately large modeled association with integration, general cognitive ability, and memory.

Marcus: So the summary is that we need to look at the organization of these weak connections because they aren't just random; they have a specific functional signature.

The paper's improvements: Ines: Now let’s talk about what the authors suggest as improvements to this concept, moving into those practical applications. They are suggesting how we can use this framework better in practice.

Marcus: They’re proposing that we should use this nonlinear weighting framework not just for prediction but also for simulating brain functional networks in a whole-cortex dynamic model, which makes the simulations more realistic.

Yuki: That’s important because simulations are a great way to see how the network behaves over time, and if you use the scaling correctly, those dynamics should look much closer to what we see in living brains.

Ines: They're also suggesting that fusing multiple filtering methods, like SIFT two and COMMIT two, is a good way to select more reliable connections while preserving those important weak ones.

Marcus: And they showed that this fused structural connectivity matrix performs better than just using COMMIT two alone in terms of how well it reveals the structure-function coupling.

Yuki: So, they're saying reliability-aware methods are helpful tools, but you still need to apply that nonlinear scaling on top of them for the best results.

Ines: And they found that this approach maintains high ICC values for both high and low reliability connectivity in their test-retest dataset, which supports the idea that weak connections remain informative even when filtered.

Marcus: That's a strong result because it shows the nonlinear weighting is robust to some of the filtering process, which means we don't lose much signal in this process.

Yuki: It’s reassuring that even after applying these rigorous filters, the functional gains from nonlinear scaling hold up, which adds confidence to using this method.

Ines: So the improvement is moving toward a framework where we can reliably characterize structural brain connectivity by integrating filtering and nonlinear scaling to get a more robust picture of function.

Conclusion: Marcus: So we’ve covered the main points of "Overlooked weak structural connections support human cognition under nonlinear connectome scaling." We've seen how the authors use a nonlinear weighting to show that those weak links matter for prediction, especially when beta is less than one point zero.

Ines: The authors conclude that these weak connections are functionally relevant because they maintain a balance between local segregation and global integration, and this influence is amplified under this nonlinear weighting.

Yuki: From a population perspective, it’s telling us that these structural features aren't just random; they are organized along systems-level and transcriptomic gradients.

Marcus: The specific pattern of weak connections linking visual/motor to limbic regions with negative gene co-expression is a key finding they want us to remember as a strong signal.

Ines: So the final message is that weak connections carry reproducible, functionally informative signals when we apply this nonlinear scaling framework. This is a way forward for studying brain structure and function together that accounts for the continuum of connection weights.

Marcus: We’re ready to move on to the next paper once we wrap up on this one, but it’s been a fascinating look at how structural biology can be used to predict cognitive traits using this new scaling logic.

Yuki: I think the insight about organizational principles is really what makes this work worth remembering for the long term.

Ines: Agreed. It's a solid piece of work showing that functional relevance isn't just about strength; it’s about context and how you measure that context through this nonlinear scaling.

State Key Laboratory for Strength and Vibration of Mechanical Structures at Jiaotong University · Department of Physics at Hong Kong Baptist University · Centre for Nonlinear Studies and Beijing-Hong Kong-Singapore Joint Centre for Nonlinear and Complex Systems (Hong Kong) Institute of Computational and Theoretical Studies at Hong Kong Baptist University · Department of Psychology at Carl von Ossietzky Universität Oldenburg · Department of Radiology and Biomedical Imaging at University of California San Francisco · Shanghai Artificial Intelligence Laboratory · School of Psychology at University of Roehampton London · Life Science Imaging Centre at Hong Kong Baptist University; Department of Physics at Zhejiang University

q-bio.NC

Submitted: 2025-05-30

Updated: 2026-11-17

Comments: 32 pages, 5 figures

Code: https://github.com/TobousRong/weak-connectivity

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 91/100

The gist: Human cognition depends on large-scale communication constrained by white-matter architecture The authors show that when tractography-derived connectivity weights are interpreted through a nonlinear…

Key concepts

Nonlinear Weighting Framework
This framework tests the idea that the functional importance of structural connections isn't linear. By applying a scaling factor $\beta$ (where $\beta < 1.0$), weak connections are amplified, meaning small absolute weights can lead to large functional effects on cognitive predictions.
Reliability-Aware Connectome
This method improves the brain map by combining two filtering techniques (SIFT2 and COMMIT2). This fusion creates a more robust structural connectome that better preserves weak connections containing important individual differences, leading to better functional simulations.
Segregation–Integration Balance
Weak connections are found to be crucial for balancing two opposing brain functions: segregation (local specialization) and integration (large-scale communication). This balance is key for cognitive abilities like general intelligence and memory, which are sensitive to these connections.

Terminology

Summary

Human cognition depends on large-scale communication constrained by white-matter architecture The authors show that when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling. p. 0

Where the paper says it

Although weak connections are abundant in mammalian connectomes, they have long been treated as noise and downweighted because of tractography uncertainty in the human brain, and their relevance to human cognition and large-scale functional organization remains unresolved. Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling.

Pruning redundant reasoning tokens

The paper investigates the role of weak structural connections in human cognition by proposing a nonlinear scaling framework for their functional influence. This framework hypothesizes that the effective functional influence of structural connectivity may follow a nonlinear scaling framework rather than a strictly linear weighting rule. p. 2

Where the paper says it

Under this assumption, weak connections may exert negligible influence on graph metrics, brain dynamics, and cognitive prediction [17, 24] because their weights are several orders of magnitude smaller than those of strong connectivity. Instead of imposing an arbitrary threshold, we therefore hypothesized that the effective functional influence of structural connectivity may follow a nonlinear scaling framework rather than a strictly linear weighting rule.

The authors test this by transforming each connection weight as w ij β, where β is an effective, target-dependent weighting parameter. p. 2
Where the paper says it

To test such a possibility, we used a phenomenological nonlinear scaling model in which each connection weight 𝑤𝑖𝑗 was transformed as 𝑤𝑖𝑗 𝛽 . Here, β is an effective, target-dependent weighting parameter rather than a universal biological constant.

When β = 1.0, the model reduces to the conventional linear assumption in previous studies. p. 2
Where the paper says it

When β = 1. 0, the model reduces to the conventional linear assumption in previous studies [17-23]; when 0.

Under this nonlinear weighting framework, weak connections could make disproportionately large contributions to brain functional organization and cognition despite their small absolute weights. p. 2
Where the paper says it

0, the transformation selectively amplifies the effective influence of weak connectivity relative to strong connectivity. Under this nonlinear weighting framework, weak connections could make disproportionately large contributions to brain functional organization and cognition despite their small absolute weights.

How the training signal is built

The authors constructed a connectome-based prediction model (CPM) to predict cognitive abilities with scaled connectivity weight w ij β. They found that general cognitive ability was predicted above chance from SC, memory and crystallized intelligence show only modest predictive associations. p. 3

Where the paper says it

We then used a connectome-based prediction model (CPM) [33] to predict cognitive abilities with scaled connectivity weight 𝑤𝑖𝑗 𝛽 (Fig. We found that general cognitive ability was predicted above chance from SC, memory and crystallized intelligence show only modest predictive associations (Fig.

More importantly, β < 1. p. 0
Where the paper says it

These effects are selective: nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain networks to enhance both large-scale integration and fine-grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities. Importantly, these effects are replicated in a reliability-aware connectome generated by integrating two post-tractography filtering methods, in which preserving weak links consistently outperforms conventional

0 appeared to generate better predictions. p. 3
Where the paper says it

More importantly, 𝛽 < 1. 0 appeared to generate better predictions.

The optimal β values were approximately β = 0. p. 3
Where the paper says it

To test this improvement while avoiding optimistic bias, we introduced an inner loop to select the optimal β inside a fully nested cross-validation in the CPM (see Materials and Methods). The optimal β values were approximately 𝛽 = 0.

703 (95% CI, 0. p. 3
Where the paper says it

The optimal β values were approximately 𝛽 = 0. 703 (95% CI, 0.

641 − 0.765) for general cognitive ability, and β = 0. p. 3
Where the paper says it

703 (95% CI, 0. 765) for general cognitive ability, and 𝛽 = 0.

602 (95% CI, 0. p. 3
Where the paper says it

765) for general cognitive ability, and 𝛽 = 0. 602 (95% CI, 0.

532 − 0.672) for memory, significantly higher than the predictions for β = 1. p. 3
Where the paper says it

602 (95% CI, 0. 672) for memory, significantly higher than the predictions for 𝛽 = 1.

0.

Where the gains come from

The findings suggest that weak connections are functionally relevant in predicting cognitive abilities under nonlinear weighting. Specifically, rewiring high-reliability weak connections resulted in lower prediction accuracy than rewiring low-reliability weak connections, highlighting the meaningful contribution of highly reliable weak connections to cognitive abilities. p. 5

Where the paper says it

We found that rewiring high-reliability weak connections resulted in lower prediction accuracy than rewiring low-reliability weak connections (Fig. 1E and S10), highlighting the meaningful contribution of highly reliable weak connections to cognitive abilities.

Furthermore, these results indicate that weak connections support both large-scale integration and fine-scale segregation by expanding the modal repertoire of brain network hierarchy. p. 0
Where the paper says it

Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling. These effects are selective: nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain networks to enhance both large-scale integration and fine-grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities.

The analysis also identified a specific class of weak connections, predominantly linking visual and motor systems with limbic regions and characterized by negative gene co-expression, which exerts a disproportionately large influence on brain function. p. 1
Where the paper says it

Finally, we show that weak connections contain functionally informative subsets organized along systems-level and transcriptomic gradients. In particular, a specific class of weak connections, predominantly linking visual and motor systems with limbic regions and characterized by negative gene co-expression, exerts a disproportionately large influence on brain function.

Organizational principles and functional heterogeneity of weak connections

The study examined the organizational principle of weak connections in terms of functional systems and gene co-expression. They found that weak connections with negative GC are mainly linked to visual/motor and limbic systems, showing a disproportionately large modeled association with brain network integration, general cognitive ability and memory. p. 1

Where the paper says it

Finally, we show that weak connections contain functionally informative subsets organized along systems-level and transcriptomic gradients. In particular, a specific class of weak connections, predominantly linking visual and motor systems with limbic regions and characterized by negative gene co-expression, exerts a disproportionately large influence on brain function.

This spatial gradient analysis suggested functional heterogeneity among weak connections. p. 6
Where the paper says it

Here, the proportion of the strongest connections was 10% (see similar results for other proportions in Fig. Cognitive prediction gains persist in the reliability-aware connectome Because noninvasive tractography is prone to false positives, particularly among weak and long-range connections [14], we next asked whether cognitive prediction gains from nonlinear weighting persist in a more reliable connectome.

p. 6
Where the paper says it

Here, the proportion of the strongest connections was 10% (see similar results for other proportions in Fig. Cognitive prediction gains persist in the reliability-aware connectome Because noninvasive tractography is prone to false positives, particularly among weak and long-range connections [14], we next asked whether cognitive prediction gains from nonlinear weighting persist in a more reliable connectome.

Weak connectivity with negative GC is more often linked visual, motor, and limbic regions, and has a disproportionately large modeled association with brain network integration, general cognitive ability and memory. p. 1
Where the paper says it

In particular, a specific class of weak connections, predominantly linking visual and motor systems with limbic regions and characterized by negative gene co-expression, exerts a disproportionately large influence on brain function. Our findings provide compelling evidence that weak connectivity plays an underestimated yet essential role in human cognition under nonlinear scaling and establish a refined framework for reliably characterizing structural brain connectivity.

Reliability-aware connectome

The authors developed a reliability-aware connectome by integrating two post-tractography filtering methods (SIFT2 [16] and COMMIT2 [3]). Fusing the connectivity weight of SIFT2 and the connectivity mask of COMMIT2 supports cognitive prediction and FC simulation while reducing likely false-positive connections. p. 6

Where the paper says it

Therefore, to reduce likely false-positive contributions and retain informative weights, we constructed a fused SC matrix (Fig. 2A) by applying COMMIT2-derived connectivity mask while retaining SIFT2 SC weight.

The fused SC had consistently higher ICCs for both high/low-reliability connectivity (ICC>0. p. 7
Where the paper says it

2B, Table S1). Moreover, fused SC had consistently higher ICCs for both high/low-reliability connectivity (ICC>0.

6 and ICC<0. p. 5
Where the paper says it

6 segregation–integration demands across cognitive abilities. Finally, in a test-retest HCP dataset with 44 individuals having two DWI scans, we calculated the intraclass correlation coefficient (ICC) to evaluate the reliability of each connectivity.

4) than COMMIT2 SC in the test-retest dataset. p. 7
Where the paper says it

6 and ICC<0. 4) than COMMIT2 SC in the test-retest dataset (Fig.

This suggests that even in the reliability-aware connectome, nonlinear weighting robustly improves the prediction of general cognitive ability and memory. p. 7
Where the paper says it

Importantly, these organizational principles were preserved in the reliability-aware connectome (fused SC), where selected weak connectivity remained enriched around hubs, and rewiring high-reliability weak connectivity continued to induce larger reductions in cognitive prediction, and the opposite effects of within- versus between-module rewiring across cognitive domains were also preserved (Fig. Therefore, even in the reliability-aware connectome, nonlinear weighting robustly improves the prediction of general cognitive ability and memory, supporting the view that weak connections carry biologically meaningful information rather than merely reflecting tractography noise.

Weak connections support segregation–integration balance

The analysis using nested spectral partition (NSP) method revealed that weak connections contribute to brain functional organization by primarily enhancing network integration. This may explain why general cognitive ability and memory, which are associated with integration and balance, may be more sensitive to weak connections with the nonlinear functioning manner (i. p. 0

Where the paper says it

Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling. These effects are selective: nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain networks to enhance both large-scale integration and fine-grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities.

e., β < 1), but crystallized intelligence that requires higher segregation is insensitive to β. p. 0
Where the paper says it

Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling. These effects are selective: nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain networks to enhance both large-scale integration and fine-grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities.

The study also found that weak connections expand the modal repertoire of brain network hierarchy by supporting both large-scale integration and fine-scale segregation. p. 0
Where the paper says it

Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling. These effects are selective: nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain networks to enhance both large-scale integration and fine-grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities.

Gene-coexpression gradient

The authors identified gene-coexpression defined subtypes of weak connections based on the sign of gene co-expression. Weak connectivity with negative GC is more often linked visual/motor and limbic systems, and has a disproportionately large modeled association with brain network integration, general cognitive ability and memory. p. 1

Where the paper says it

In particular, a specific class of weak connections, predominantly linking visual and motor systems with limbic regions and characterized by negative gene co-expression, exerts a disproportionately large influence on brain function. Our findings provide compelling evidence that weak connectivity plays an underestimated yet essential role in human cognition under nonlinear scaling and establish a refined framework for reliably characterizing structural brain connectivity.

This spatial gradient analysis suggested functional heterogeneity among weak connections.

Conclusion

The proposed framework suggests that weak connections are functionally relevant to cognitive abilities by integrating different systems to maintain a balance between local segregation and global integration, with the functional influence amplified under nonlinear weighting. The authors conclude that weak connections appear to carry reproducible, functionally-informative signal. p. 5

Where the paper says it

Overall, our results indicate that under nonlinear weighting, weak connections are functionally relevant in predicting cognitive abilities, suggesting that their contribution should not be attributed solely to noise. In fact, although nonlinear weighting may also increase the relative influence of noisy connections, the cognitive predictions appear to be highly nontrivial because the CPM showed substantial robustness to noise, which was explicitly tested by introducing more false positives (Fig.

Materials and Methods

The primary dataset was obtained from the WU-Minn Human Connectome Project (HCP) Young Adult database. Diffusion MRI data processing involved reconstructing whole-cortex probabilistic streamlines using the iFOD2 algorithm, producing 10 million streamlines with backtracking enabled. p. 3

Where the paper says it

Finally, we examined the organizational principle of weak connections in the brain in terms of functional systems and gene co-expression, and investigated the heterogeneous functions of weak connectivity in relation to brain function and cognitive abilities. Results Nonlinear weighting reveals functional relevance of weak connectivity to cognitive abilities We used probabilistic tractography [30] to reconstruct streamlines and construct whole-cortex basic SC networks (N=360 regions, binary density: 56.

The connectome-based prediction model (CPM) used a nonlinear kernel regression where an exponent scale parameter β was applied to all SC. p. 3
Where the paper says it

These four latent abilities were selected to probe different functional demands on network segregation (required by crystallized intelligence and processing speed), integration (required by general cognitive ability) and their balance (required by memory) [32]. We then used a connectome-based prediction model (CPM) [33] to predict cognitive abilities with scaled connectivity weight 𝑤𝑖𝑗 𝛽 (Fig.

Structural connectome modes

The structural modes were obtained through the construction of the Laplacian matrix and subsequent eigenmode calculation. The active size of structural modes was calculated by defining it as the number of regions with squared eigenvector component values larger than a threshold. p. 4

Where the paper says it

We examined whether weak connections selected by CPM were topologically organized in terms of topological properties First, we calculated the weighted degree of regions in the group-averaged SC network, and selected those regions in the top 10% degree as structural hubs (see similar results for functional hubs in Fig. Rather than explicitly defining weak connectivity using a threshold, we fixed the proportion of the strongest connections (i.

The analysis showed that weak connections support both large-scale integration and fine-scale segregation, thereby expanding the modal repertoire of brain network hierarchy. p. 0
Where the paper says it

Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling. These effects are selective: nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain networks to enhance both large-scale integration and fine-grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities.

Alignment of structural modes

The authors performed alignment by minimizing the angle variance between thresholded network modes and reference modes from original full individual SC networks. This alignment procedure was necessary to precisely study the effect of weak connections on functional activity across different network densities. p. 6

Where the paper says it

SIFT2 mainly decreases the weight of weak connections, but does not remove the links (Fig. COMMIT2 introduces biological constraints on the formation of bundles to reduce false-positives [3] while suppressing and removing weak connections to generate different network densities by varying the model parameter 𝜆 (Fig.

Active size of structural modes

The normalized active sizes for low-order modes and high-order mode were calculated separately after aligning the modes. The results showed that weak connections support both large-scale integration and fine-scale segregation, thereby expanding the modal repertoire of brain network hierarchy. p. 0

Where the paper says it

Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling. These effects are selective: nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain networks to enhance both large-scale integration and fine-grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities.

The gene expression data were mapped to 360 regions in the MMP atlas. The Pearson correlation between the gene expression profiles of regions i and j was calculated to represent the gene coexpression (GC) between these two regions. p. 1

Where the paper says it

Finally, we show that weak connections contain functionally informative subsets organized along systems-level and transcriptomic gradients. In particular, a specific class of weak connections, predominantly linking visual and motor systems with limbic regions and characterized by negative gene co-expression, exerts a disproportionately large influence on brain function.

The GC matrices were transformed into cosine distance matrices, and then classical multidimensional scaling (CMS) method was applied to obtain the coordinates of regions in the low-dimensional space using cmdscale. p. 9
Where the paper says it

, minimizing FC distance), yielding an optimum around β =0. To test this result while avoiding optimistic bias, we generated two FC matrices by concatenating two fMRI time series in each individual, using one for model fitting and the other for testing.

Statistical analysis

The statistical analysis used the two-sample t-test and paired t-test with Matlab software. The FDR correction was performed for the exploratory analyses. p. 4

Where the paper says it

This enrichment suggests that selected weak connections may provide hub-to-hub parallel pathways relevant to cognitive prediction, consistent with a recent finding [34]. Then, using the optimal β identified in the trained models, we performed a series of degreepreserving rewiring analyses to evaluate the contribution of weak connectivity to cognitive prediction, i.

The permutation test of Moran spectral randomization was based on BrainSpace toolbox. p. 9
Where the paper says it

We next asked whether the functional gains associated with nonlinear weighting of weak connectivity persist after reliability-aware control. We used the nested spectral partition (NSP) method to quantify segregation-integration balance [32], i.

Acknowledgments

This work was partially supported by the STI 2030-Major Projects (2022ZD0208500), the National Natural Science Foundation of China (Nos. 12132012, 32541016, 12272294, and the Hong Kong Research Grant Council. p. 0

Where the paper says it

State Key Laboratory for Strength and Vibration of Mechanical Structures, School of Aerospace Engineering, Xi’an Jiaotong University, Xi’an 710049, China 2. Department of Physics, Hong Kong Baptist University, Kowloon Tong, Hong Kong 3.

Competing interests

Authors declare that they have no competing interests.

Data Availability Statement

The MRI and cognitive datasets are available at http://www.humanconnectome.org/study/hcpyoung-adult. p. 1

Where the paper says it

Diffusion MRI tractography provides the only noninvasive approach for reconstructing these pathways in vivo, commonly referred to as structural connectivity (SC) [3, 4]. Despite well-known limitations [3-6], including false positives and uncertainty in mapping white-matter pathways, tractography has substantially advanced the study of brain function, cognition, and neuropsychiatric disorders [4, 7-10], and serves as a foundational input for large-scale brain modeling platforms such as The Virtual Brain [11] and Digital Twin Brain [12].

The code used for this study is available at https://github. p. 1
Where the paper says it

Diffusion MRI tractography provides the only noninvasive approach for reconstructing these pathways in vivo, commonly referred to as structural connectivity (SC) [3, 4]. Despite well-known limitations [3-6], including false positives and uncertainty in mapping white-matter pathways, tractography has substantially advanced the study of brain function, cognition, and neuropsychiatric disorders [4, 7-10], and serves as a foundational input for large-scale brain modeling platforms such as The Virtual Brain [11] and Digital Twin Brain [12].

com/TobousRong/weak-connectivity. p. 0
Where the paper says it

Although weak connections are abundant in mammalian connectomes, they have long been treated as noise and downweighted because of tractography uncertainty in the human brain, and their relevance to human cognition and large-scale functional organization remains unresolved. Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling.

The deep learning code to predict SC from FC is available at https://github. p. 9
Where the paper says it

300, p < 0. When using a deep learning model to predict fused SC from FC, the correlation between predicted and empirical SC peaked at 𝛽 = 0.

com/qidianzl/RecoveringBrain-Structure-Network-Using-Functional-Connectivity. p. 4
Where the paper says it

We examined whether weak connections selected by CPM were topologically organized in terms of topological properties First, we calculated the weighted degree of regions in the group-averaged SC network, and selected those regions in the top 10% degree as structural hubs (see similar results for functional hubs in Fig. Rather than explicitly defining weak connectivity using a threshold, we fixed the proportion of the strongest connections (i.

References

  1. Bullmore, E. and O. Sporns, Complex brain networks: graph theoretical analysis of structural and functional systems. p. 2

    Where the paper says it

    3 indeed had little effect on graph-theoretical properties of brain networks [17], and removing the weakest ∼50% of connections had negligible effects on the neural signal-propagation ratio in a rat-brain dynamic model [24]. However, invasive tract-tracing studies across mammalian brains consistently reveal dense cortical architectures with connectivity weight spanning several orders of magnitude (e.

  2. Gollo, L.L., et al., Fragility and volatility of structural hubs in the human connectome. p. 0

    Where the paper says it

    1 Overlooked weak structural connections support human cognition under nonlinear connectome scaling Rong Wang1#,*, Zhao Chang2,3# , Xuechun Liu1 , Daniel Kristanto4 , Étienne Gérard Guy Gartner2,3, Xinyang Liu5 , Mianxin Liu6 , Ying Wu1 , Ming Lui7 , Changsong Zhou2,3,,8,9* 1. State Key Laboratory for Strength and Vibration of Mechanical Structures, School of Aerospace Engineering, Xi’an Jiaotong University, Xi’an 710049, China 2.

  3. Schiavi, S., et al., A new method for accurate in vivo mapping of human brain connections using microstructural and anatomical information. p. 1

    Where the paper says it

    Diffusion MRI tractography provides the only noninvasive approach for reconstructing these pathways in vivo, commonly referred to as structural connectivity (SC) [3, 4]. Despite well-known limitations [3-6], including false positives and uncertainty in mapping white-matter pathways, tractography has substantially advanced the study of brain function, cognition, and neuropsychiatric disorders [4, 7-10], and serves as a foundational input for large-scale brain modeling platforms such as The Virtual Brain [11] and Digital Twin Brain [12].

  4. Maier-Hein, K.H., et al., The challenge of mapping the human connectome based on diffusion tractography. p. 1

    Where the paper says it

    Diffusion MRI tractography provides the only noninvasive approach for reconstructing these pathways in vivo, commonly referred to as structural connectivity (SC) [3, 4]. Despite well-known limitations [3-6], including false positives and uncertainty in mapping white-matter pathways, tractography has substantially advanced the study of brain function, cognition, and neuropsychiatric disorders [4, 7-10], and serves as a foundational input for large-scale brain modeling platforms such as The Virtual Brain [11] and Digital Twin Brain [12].

  5. Jones, D.K. and M. Cercignani, Twenty‐five pitfalls in the analysis of diffusion MRI data. p. 1

    Where the paper says it

    Rather than being well described as sparse binary graphs, the brain structural connectomes exhibit heavy-tailed weight distributions spanning several orders of magnitude, with numerous weak connections embedded alongside the strong pathways. Diffusion MRI tractography provides the only noninvasive approach for reconstructing these pathways in vivo, commonly referred to as structural connectivity (SC) [3, 4].

  6. Jones, D.K., T.R. Knösche, and R. Turner, White matter integrity, fiber count, and other fallacies: the do's and don'ts of diffusion MRI.

  7. Sporns, O., The human connectome: a complex network. p. 0

    Where the paper says it

    1 Overlooked weak structural connections support human cognition under nonlinear connectome scaling Rong Wang1#,*, Zhao Chang2,3# , Xuechun Liu1 , Daniel Kristanto4 , Étienne Gérard Guy Gartner2,3, Xinyang Liu5 , Mianxin Liu6 , Ying Wu1 , Ming Lui7 , Changsong Zhou2,3,,8,9* 1. State Key Laboratory for Strength and Vibration of Mechanical Structures, School of Aerospace Engineering, Xi’an Jiaotong University, Xi’an 710049, China 2.

  8. Fornito, A., A. Zalesky, and M. Breakspear, The connectomics of brain disorders. p. 1

    Where the paper says it

    Diffusion MRI tractography provides the only noninvasive approach for reconstructing these pathways in vivo, commonly referred to as structural connectivity (SC) [3, 4]. Despite well-known limitations [3-6], including false positives and uncertainty in mapping white-matter pathways, tractography has substantially advanced the study of brain function, cognition, and neuropsychiatric disorders [4, 7-10], and serves as a foundational input for large-scale brain modeling platforms such as The Virtual Brain [11] and Digital Twin Brain [12].

  9. Lodygensky, G.A., et al., Neuroimaging of cortical development and brain connectivity in human newborns and animal models. p. 1

    Where the paper says it

    Our findings provide compelling evidence that weak connectivity plays an underestimated yet essential role in human cognition under nonlinear scaling and establish a refined framework for reliably characterizing structural brain connectivity. Keywords: structural connectivity, cognition, large-scale dynamic model, connectome prediction model, tractography Introduction Human cognition emerges from large-scale interactions constrained by the brain structural connectome, in which cortical and subcortical regions are linked through white-matter pathways [1, 2].

  10. Ardesch, D.J., et al., Evolutionary expansion of connectivity between multimodal association areas in the human brain compared with chimpanzees. p. 0

    Where the paper says it

    Although weak connections are abundant in mammalian connectomes, they have long been treated as noise and downweighted because of tractography uncertainty in the human brain, and their relevance to human cognition and large-scale functional organization remains unresolved. Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling.

  11. Ritter, P., et al., The virtual brain integrates computational modeling and multimodal neuroimaging. p. 1

    Where the paper says it

    Diffusion MRI tractography provides the only noninvasive approach for reconstructing these pathways in vivo, commonly referred to as structural connectivity (SC) [3, 4]. Despite well-known limitations [3-6], including false positives and uncertainty in mapping white-matter pathways, tractography has substantially advanced the study of brain function, cognition, and neuropsychiatric disorders [4, 7-10], and serves as a foundational input for large-scale brain modeling platforms such as The Virtual Brain [11] and Digital Twin Brain [12].

  12. Lu, W., et al., Imitating and exploring the human brain's resting and task-performing states via brain computing: scaling and architecture.

  13. Betzel, R.F. and D.S. Bassett, Specificity and robustness of long-distance connections in weighted, interareal connectomes. p. 0

    Where the paper says it

    hk ABSTRACT Human cognition depends on large-scale communication constrained by white-matter architecture. Although weak connections are abundant in mammalian connectomes, they have long been treated as noise and downweighted because of tractography uncertainty in the human brain, and their relevance to human cognition and large-scale functional organization remains unresolved.

  14. Jones, D.K., Studying connections in the living human brain with diffusion MRI. p. 1

    Where the paper says it

    Rather than being well described as sparse binary graphs, the brain structural connectomes exhibit heavy-tailed weight distributions spanning several orders of magnitude, with numerous weak connections embedded alongside the strong pathways. Diffusion MRI tractography provides the only noninvasive approach for reconstructing these pathways in vivo, commonly referred to as structural connectivity (SC) [3, 4].

  15. Reveley, C., et al., Superficial white matter fiber systems impede detection of long-range cortical connections in diffusion MRI. p. 1

    Where the paper says it

    Diffusion MRI tractography provides the only noninvasive approach for reconstructing these pathways in vivo, commonly referred to as structural connectivity (SC) [3, 4]. Despite well-known limitations [3-6], including false positives and uncertainty in mapping white-matter pathways, tractography has substantially advanced the study of brain function, cognition, and neuropsychiatric disorders [4, 7-10], and serves as a foundational input for large-scale brain modeling platforms such as The Virtual Brain [11] and Digital Twin Brain [12].

  16. Smith, R.E., et al., SIFT2: Enabling dense quantitative assessment of brain white matter connectivity using streamlines tractography. p. 1

    Where the paper says it

    Despite well-known limitations [3-6], including false positives and uncertainty in mapping white-matter pathways, tractography has substantially advanced the study of brain function, cognition, and neuropsychiatric disorders [4, 7-10], and serves as a foundational input for large-scale brain modeling platforms such as The Virtual Brain [11] and Digital Twin Brain [12]. However, the functional relevance of the widespread weak connectivity within these dense connectomes remains poorly understood.

  17. Civier, O., et al., Is removal of weak connections necessary for graph-theoretical analysis of dense weighted structural connectomes from diffusion MRI? p. 1

    Where the paper says it

    Rather than being well described as sparse binary graphs, the brain structural connectomes exhibit heavy-tailed weight distributions spanning several orders of magnitude, with numerous weak connections embedded alongside the strong pathways. Diffusion MRI tractography provides the only noninvasive approach for reconstructing these pathways in vivo, commonly referred to as structural connectivity (SC) [3, 4].

  18. Li, Y., et al., Brain anatomical network and intelligence. p. 3

    Where the paper says it

    We first extracted individual estimates of crystallized intelligence (cry), processing speed (spd), general cognitive ability (g) and memory (mem) through a structural equation model (SEM) (Fig. These four latent abilities were selected to probe different functional demands on network segregation (required by crystallized intelligence and processing speed), integration (required by general cognitive ability) and their balance (required by memory) [32].

  19. Li, C., et al., Impaired topological architecture of brain structural networks in idiopathic Parkinson’s disease: a DTI study. p. 1

    Where the paper says it

    Diffusion MRI tractography provides the only noninvasive approach for reconstructing these pathways in vivo, commonly referred to as structural connectivity (SC) [3, 4]. Despite well-known limitations [3-6], including false positives and uncertainty in mapping white-matter pathways, tractography has substantially advanced the study of brain function, cognition, and neuropsychiatric disorders [4, 7-10], and serves as a foundational input for large-scale brain modeling platforms such as The Virtual Brain [11] and Digital Twin Brain [12].

  20. Huang, H., et al., Development of human brain structural networks through infancy and childhood. p. 1

    Where the paper says it

    In particular, a specific class of weak connections, predominantly linking visual and motor systems with limbic regions and characterized by negative gene co-expression, exerts a disproportionately large influence on brain function. Our findings provide compelling evidence that weak connectivity plays an underestimated yet essential role in human cognition under nonlinear scaling and establish a refined framework for reliably characterizing structural brain connectivity.

  21. Zheng, M., et al., Geometric renormalization unravels self-similarity of the multiscale human connectome. p. 0

    Where the paper says it

    1 Overlooked weak structural connections support human cognition under nonlinear connectome scaling Rong Wang1#,*, Zhao Chang2,3# , Xuechun Liu1 , Daniel Kristanto4 , Étienne Gérard Guy Gartner2,3, Xinyang Liu5 , Mianxin Liu6 , Ying Wu1 , Ming Lui7 , Changsong Zhou2,3,,8,9* 1. State Key Laboratory for Strength and Vibration of Mechanical Structures, School of Aerospace Engineering, Xi’an Jiaotong University, Xi’an 710049, China 2.

  22. Seguin, C., Y. Tian, and A. Zalesky, Network communication models improve the behavioral and functional predictive utility of the human structural connectome. p. 0

    Where the paper says it

    hk ABSTRACT Human cognition depends on large-scale communication constrained by white-matter architecture. Although weak connections are abundant in mammalian connectomes, they have long been treated as noise and downweighted because of tractography uncertainty in the human brain, and their relevance to human cognition and large-scale functional organization remains unresolved.

  23. Vézquez-Rodríguez, B., et al., Signal propagation via cortical hierarchies. p. 2

    Where the paper says it

    3 indeed had little effect on graph-theoretical properties of brain networks [17], and removing the weakest ∼50% of connections had negligible effects on the neural signal-propagation ratio in a rat-brain dynamic model [24]. However, invasive tract-tracing studies across mammalian brains consistently reveal dense cortical architectures with connectivity weight spanning several orders of magnitude (e.

  24. Joglekar, M.R., et al., Inter-areal Balanced Amplification Enhances Signal Propagation in a Large-Scale Circuit Model of the Primate Cortex. p. 2

    Where the paper says it

    3 indeed had little effect on graph-theoretical properties of brain networks [17], and removing the weakest ∼50% of connections had negligible effects on the neural signal-propagation ratio in a rat-brain dynamic model [24]. However, invasive tract-tracing studies across mammalian brains consistently reveal dense cortical architectures with connectivity weight spanning several orders of magnitude (e.

  25. Gămănuţ, R., et al., The Mouse Cortical Connectome, Characterized by an Ultra-Dense Cortical Graph, Maintains Specificity by Distinct Connectivity Profiles. p. 2

    Where the paper says it

    However, invasive tract-tracing studies across mammalian brains consistently reveal dense cortical architectures with connectivity weight spanning several orders of magnitude (e. , 97% for 47 areas in mouse brains) [25-28].

  26. Oh, S.W., et al., A mesoscale connectome of the mouse brain. p. 0

    Where the paper says it

    These effects are selective: nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain networks to enhance both large-scale integration and fine-grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities. Importantly, these effects are replicated in a reliability-aware connectome generated by integrating two post-tractography filtering methods, in which preserving weak links consistently outperforms conventional

  27. Markov, N.T., et al., A weighted and directed interareal connectivity matrix for macaque cerebral cortex. p. 4

    Where the paper says it

    We examined whether weak connections selected by CPM were topologically organized in terms of topological properties First, we calculated the weighted degree of regions in the group-averaged SC network, and selected those regions in the top 10% degree as structural hubs (see similar results for functional hubs in Fig. Rather than explicitly defining weak connectivity using a threshold, we fixed the proportion of the strongest connections (i.

  28. Theodoni, P., et al., Structural attributes and principles of the neocortical connectome in the marmoset monkey. p. 7

    Where the paper says it

    2B and S13). Importantly, these organizational principles were preserved in the reliability-aware connectome (fused SC), where selected weak connectivity remained enriched around hubs, and rewiring high-reliability weak connectivity continued to induce larger reductions in cognitive prediction, and the opposite effects of within- versus between-module rewiring across cognitive domains were also preserved (Fig.

  29. Markov, N.T., et al., Cortical high-density counterstream architectures. p. 2

    Where the paper says it

    3 indeed had little effect on graph-theoretical properties of brain networks [17], and removing the weakest ∼50% of connections had negligible effects on the neural signal-propagation ratio in a rat-brain dynamic model [24]. However, invasive tract-tracing studies across mammalian brains consistently reveal dense cortical architectures with connectivity weight spanning several orders of magnitude (e.

  30. Tournier, J.-D., et al., MRtrix3: A fast, flexible and open software framework for medical image processing and visualisation. p. 0

    Where the paper says it

    Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling. These effects are selective: nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain networks to enhance both large-scale integration and fine-grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities.

  31. Chen, Y., et al., A Large-Scale High-Density Weighted Structural Connectome of the Macaque Brain Acquired by Predicting Missing Links. p. 0

    Where the paper says it

    These effects are selective: nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain networks to enhance both large-scale integration and fine-grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities. Importantly, these effects are replicated in a reliability-aware connectome generated by integrating two post-tractography filtering methods, in which preserving weak links consistently outperforms conventional

  32. Wang, R., et al., Segregation, integration and balance of large-scale resting brain networks configure different cognitive abilities. p. 0

    Where the paper says it

    Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling. These effects are selective: nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain networks to enhance both large-scale integration and fine-grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities.

  33. Pedregosa, F., et al., Scikit-learn: Machine learning in Python. p. 2

    Where the paper says it

    Here, using the HCP (n=999) and UCLA datasets (n = 122), we constructed SC matrices with multiple tractography and filtering pipelines. First, we constructed a connectome-based machine learning model to investigate whether nonlinear weighting reveals the functional relevance of weak connectivity to cognitive abilities and whether the functional relevance persists after accounting for likely tractography false positives.

  34. Fakhar, K., et al., A general framework for characterizing optimal communication in brain networks. p. 0

    Where the paper says it

    Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling. These effects are selective: nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain networks to enhance both large-scale integration and fine-grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities.

  35. Schilling, K.G., et al., Tractography dissection variability: What happens when 42 groups dissect 14 white matter bundles on the same dataset. p. 0

    Where the paper says it

    hk ABSTRACT Human cognition depends on large-scale communication constrained by white-matter architecture. Although weak connections are abundant in mammalian connectomes, they have long been treated as noise and downweighted because of tractography uncertainty in the human brain, and their relevance to human cognition and large-scale functional organization remains unresolved.

  36. Wang, R., et al., Hierarchical connectome modes and critical state jointly maximize human brain functional diversity. p. 0

    Where the paper says it

    1 Overlooked weak structural connections support human cognition under nonlinear connectome scaling Rong Wang1#,*, Zhao Chang2,3# , Xuechun Liu1 , Daniel Kristanto4 , Étienne Gérard Guy Gartner2,3, Xinyang Liu5 , Mianxin Liu6 , Ying Wu1 , Ming Lui7 , Changsong Zhou2,3,,8,9* 1. State Key Laboratory for Strength and Vibration of Mechanical Structures, School of Aerospace Engineering, Xi’an Jiaotong University, Xi’an 710049, China 2.

  37. Atasoy, S., I. Donnelly, and J. Pearson, Human brain networks function in connectome-specific harmonic waves. p. 1

    Where the paper says it

    In particular, a specific class of weak connections, predominantly linking visual and motor systems with limbic regions and characterized by negative gene co-expression, exerts a disproportionately large influence on brain function. Our findings provide compelling evidence that weak connectivity plays an underestimated yet essential role in human cognition under nonlinear scaling and establish a refined framework for reliably characterizing structural brain connectivity.

  38. Yeo, B.T., et al., The organization of the human cerebral cortex estimated by intrinsicfunctional connectivity. p. 0

    Where the paper says it

    Although weak connections are abundant in mammalian connectomes, they have long been treated as noise and downweighted because of tractography uncertainty in the human brain, and their relevance to human cognition and large-scale functional organization remains unresolved. Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling.

  39. Chen, C.-H., et al., Hierarchical Genetic Organization of Human Cortical Surface Area. p. 1

    Where the paper says it

    Our findings provide compelling evidence that weak connectivity plays an underestimated yet essential role in human cognition under nonlinear scaling and establish a refined framework for reliably characterizing structural brain connectivity. Keywords: structural connectivity, cognition, large-scale dynamic model, connectome prediction model, tractography Introduction Human cognition emerges from large-scale interactions constrained by the brain structural connectome, in which cortical and subcortical regions are linked through white-matter pathways [1, 2].

  40. Burt, J.B., et al., Hierarchy of transcriptomic specialization across human cortex captured by structural neuroimaging topography. p. 3

    Where the paper says it

    Results Nonlinear weighting reveals functional relevance of weak connectivity to cognitive abilities We used probabilistic tractography [30] to reconstruct streamlines and construct whole-cortex basic SC networks (N=360 regions, binary density: 56. SC weights spans several orders of magnitude across individuals (Fig.

  41. Margulies, D.S., et al., Situating the default-mode network along a principal gradient of macroscale cortical organization.

  42. Tian, Y., et al., Topographic organization of the human subcortex unveiled with functional connectivity gradients. p. 0

    Where the paper says it

    Although weak connections are abundant in mammalian connectomes, they have long been treated as noise and downweighted because of tractography uncertainty in the human brain, and their relevance to human cognition and large-scale functional organization remains unresolved. Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling.

  43. Rubinov, M. and O. Sporns, Complex network measures of brain connectivity: Uses and interpretations. p. 2

    Where the paper says it

    Therefore, the critical question is not only whether tractographyderived weak connections are reliable, but under what weighting assumptions they become functionally relevant to brain function and cognitive abilities. A common implicit assumption in network neuroscience is that the functional influence of structural connectivity scales linearly with tractography weight.

  44. Roberts, J.A., et al., Consistency-based thresholding of the human connectome. p. 2

    Where the paper says it

    Here, using the HCP (n=999) and UCLA datasets (n = 122), we constructed SC matrices with multiple tractography and filtering pipelines. First, we constructed a connectome-based machine learning model to investigate whether nonlinear weighting reveals the functional relevance of weak connectivity to cognitive abilities and whether the functional relevance persists after accounting for likely tractography false positives.

  45. Betzel, R.F., et al., Distance-dependent consensus thresholds for generating group-representative structural brain networks. p. 1

    Where the paper says it

    A major challenge is that weak connectivity occupies an ambiguous position between biological organization and tractography uncertainty [13-15]. Diffusion MRI–derived weak connections are also the most vulnerable to false positives, distance-dependent attenuation, and crossing-fiber uncertainty [3, 14, 15].

  46. Friedrich, M., et al., Whole-brain structural connectivity predicts cognitive deficits in pretreated glioma patients. p. 3

    Where the paper says it

    Finally, we examined the organizational principle of weak connections in the brain in terms of functional systems and gene co-expression, and investigated the heterogeneous functions of weak connectivity in relation to brain function and cognitive abilities. Results Nonlinear weighting reveals functional relevance of weak connectivity to cognitive abilities We used probabilistic tractography [30] to reconstruct streamlines and construct whole-cortex basic SC networks (N=360 regions, binary density: 56.

  47. Yoo, C., S. Park, and M.J. Kim, Structural connectome-based prediction of trait anxiety. p. 3

    Where the paper says it

    These four latent abilities were selected to probe different functional demands on network segregation (required by crystallized intelligence and processing speed), integration (required by general cognitive ability) and their balance (required by memory) [32]. We then used a connectome-based prediction model (CPM) [33] to predict cognitive abilities with scaled connectivity weight 𝑤𝑖𝑗 𝛽 (Fig.

  48. Li, X., et al., Topological reorganization after partial auditory deprivation—a structural connectivity study in singlesided deafness. p. 1

    Where the paper says it

    Diffusion MRI tractography provides the only noninvasive approach for reconstructing these pathways in vivo, commonly referred to as structural connectivity (SC) [3, 4]. Despite well-known limitations [3-6], including false positives and uncertainty in mapping white-matter pathways, tractography has substantially advanced the study of brain function, cognition, and neuropsychiatric disorders [4, 7-10], and serves as a foundational input for large-scale brain modeling platforms such as The Virtual Brain [11] and Digital Twin Brain [12].

  49. Goulas, A., A. Schaefer, and D.S. Margulies, The strength of weak connections in the macaque cortico-cortical network. p. 0

    Where the paper says it

    1 Overlooked weak structural connections support human cognition under nonlinear connectome scaling Rong Wang1#,*, Zhao Chang2,3# , Xuechun Liu1 , Daniel Kristanto4 , Étienne Gérard Guy Gartner2,3, Xinyang Liu5 , Mianxin Liu6 , Ying Wu1 , Ming Lui7 , Changsong Zhou2,3,,8,9* 1. State Key Laboratory for Strength and Vibration of Mechanical Structures, School of Aerospace Engineering, Xi’an Jiaotong University, Xi’an 710049, China 2.

  50. Simola, J., et al., Genetic polymorphisms in COMT and BDNF influence synchronization dynamics of human neuronal oscillations. p. 2

    Where the paper says it

    A common implicit assumption in network neuroscience is that the functional influence of structural connectivity scales linearly with tractography weight. Under this assumption, weak connections may exert negligible influence on graph metrics, brain dynamics, and cognitive prediction [17, 24] because their weights are several orders of magnitude smaller than those of strong connectivity.

  51. Vezoli, J., et al., Cortical hierarchy, dual counterstream architecture and the importance of top-down generative networks. p. 2

    Where the paper says it

    3 indeed had little effect on graph-theoretical properties of brain networks [17], and removing the weakest ∼50% of connections had negligible effects on the neural signal-propagation ratio in a rat-brain dynamic model [24]. However, invasive tract-tracing studies across mammalian brains consistently reveal dense cortical architectures with connectivity weight spanning several orders of magnitude (e.

  52. Latawiec, D., K.A.C. Martin, and V. Meskenaite, Termination of the geniculocortical projection in the striate cortex of macaque monkey: A quantitative immunoelectron microscopic study.

  53. Scholl, B., et al., Cortical response selectivity derives from strength in numbers of synapses. p. 0

    Where the paper says it

    1 Overlooked weak structural connections support human cognition under nonlinear connectome scaling Rong Wang1#,*, Zhao Chang2,3# , Xuechun Liu1 , Daniel Kristanto4 , Étienne Gérard Guy Gartner2,3, Xinyang Liu5 , Mianxin Liu6 , Ying Wu1 , Ming Lui7 , Changsong Zhou2,3,,8,9* 1. State Key Laboratory for Strength and Vibration of Mechanical Structures, School of Aerospace Engineering, Xi’an Jiaotong University, Xi’an 710049, China 2.

  54. Yang, Z., J. Liang, and C. Zhou, Critical Avalanches in Excitation-Inhibition Balanced Networks Reconcile Response Reliability with Sensitivity for Optimal Neural Representation. p. 2

    Where the paper says it

    3 indeed had little effect on graph-theoretical properties of brain networks [17], and removing the weakest ∼50% of connections had negligible effects on the neural signal-propagation ratio in a rat-brain dynamic model [24]. However, invasive tract-tracing studies across mammalian brains consistently reveal dense cortical architectures with connectivity weight spanning several orders of magnitude (e.

  55. Baum, G.L., et al., Modular Segregation of Structural Brain Networks Supports the Development of Executive Function in Youth. p. 0

    Where the paper says it

    Across multiple datasets and tractography pipelines, we show that, when tractography-derived connectivity weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive prediction, functional-connectivity simulation, and structure–function coupling. These effects are selective: nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain networks to enhance both large-scale integration and fine-grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities.

Discussion

The authors suggest that weak connections are functionally relevant to cognitive abilities by integrating different systems to maintain a balance between local segregation and global integration, with the functional influence amplified under nonlinear weighting. They argue that weak connections appear to carry reproducible, functionally-informative signal rather than being purely noise. p. 5

Where the paper says it

Overall, our results indicate that under nonlinear weighting, weak connections are functionally relevant in predicting cognitive abilities, suggesting that their contribution should not be attributed solely to noise. In fact, although nonlinear weighting may also increase the relative influence of noisy connections, the cognitive predictions appear to be highly nontrivial because the CPM showed substantial robustness to noise, which was explicitly tested by introducing more false positives (Fig.

Nonlinear weighting of structural connectivity

The authors propose a working hypothesis that the functional influence of structural connectivity may follow a nonlinear scaling framework rather than a strictly linear weighting rule. Under this theoretical framework, they found that weak connections contribute to more accurate predictions of cognitive abilities, a stronger relationship of brain structural and functional connectivity and a more realistic simulation of brain functional networks in the whole-cortex dynamic model. p. 3

Where the paper says it

Finally, we examined the organizational principle of weak connections in the brain in terms of functional systems and gene co-expression, and investigated the heterogeneous functions of weak connectivity in relation to brain function and cognitive abilities. Results Nonlinear weighting reveals functional relevance of weak connectivity to cognitive abilities We used probabilistic tractography [30] to reconstruct streamlines and construct whole-cortex basic SC networks (N=360 regions, binary density: 56.

Several mechanisms could plausibly contribute to such nonlinear amplification. p. 2
Where the paper says it

Under this assumption, weak connections may exert negligible influence on graph metrics, brain dynamics, and cognitive prediction [17, 24] because their weights are several orders of magnitude smaller than those of strong connectivity. Instead of imposing an arbitrary threshold, we therefore hypothesized that the effective functional influence of structural connectivity may follow a nonlinear scaling framework rather than a strictly linear weighting rule.

Fusing multiple filtering methods is a promising way to select more reliable connections, especially preserving weak connections that contain impor…

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Nonlinear weighting framework for structural connectivity can be used to improve predictions of general cognitive ability and memory by transforming connection weights as w ij to beta, with values less than 1.0 amplifying weak connection influence over strong connections.

  2. A reliability-aware connectome generated by integrating two post-tractography filtering methods, SIFT2 and COMMIT2, consistently outperforms conventional thresholding strategies in preserving functional relevance of weak links.

  3. The improved AI system can predict cognitive abilities with higher accuracy for general cognitive ability and memory by using the optimal beta values identified through a fully nested cross-validation within a connectome-based prediction model.

  4. The system can identify functionally informative subsets of weak connections organized along systems-level and transcriptomic gradients, specifically those linking visual and motor systems with limbic regions and characterized by negative gene co-expression.

  5. The AI can maintain or improve functional gains associated with nonlinear weighting even when the connectome is filtered using reliability-aware approaches like fused SC matrices, showing that fused SC had consistently higher ICCs for both high/low-reliability connectivity than COMMIT2 SC.

  6. The system can support brain functional organization by expanding the modal repertoire of brain network hierarchy, as weak connections contribute to both large-scale integration and fine-scale segregation across structural modes.

  7. The AI can differentiate the functional roles of weak connections based on gene co-expression, identifying that weak connectivity with negative GC is more often linked visual/motor and limbic regions which has a disproportionately large modeled association with brain network integration, general cognitive ability and memory.

Abstract

Human cognition depends on large scale communication constrained by white matter architecture. Although weak connections are abundant in mammalian connectomes, they have long been treated as noise and downweighted because of tractography uncertainty in the human brain, and their relevance to human cognition and large scale functional organization remains unresolved. Across multiple datasets and tractography pipelines, we show that, when tractography derived connection weights are interpreted through a nonlinear weighting framework, weak connections make measurable contributions to cognitive predictions, functional connectivity simulation, and structure function coupling. These effects are selective, nonlinear weighting improves the prediction of general cognitive ability and memory more than that of crystallized intelligence or processing speed, consistent with the notion that weak connections preferentially expand the modal repertoire of brain structural networks to enhance both large scale integration and fine grained segregation, thereby supporting the functional balance essential for diverse cognitive abilities. Importantly, these effects are replicated in a reliability aware connectome generated by integrating two post tractography filtering methods, in which preserving weak links consistently outperforms conventional thresholding strategy in structure function coupling and model based simulation. Finally, we show that weak connections contain functional information organized along system level and gene expression gradients. In particular, weak connections with negative gene coexpression predominantly link visual, motor with limbic regions and exert a disproportionately large influence on brain function than those with positive gene coexpression.

Related papers