Critical Mass: Structure-Disjoint Splits Drift With the Libraries They Are Built On
Researchers optimized a method for splitting molecular libraries into structure-disjoint folds using Maximum Common Edge Subgraph (MCES) distance at a threshold of 10. The optimization made it computationally feasible for large public libraries, revealing that the resulting holdout set shrinks and becomes systematically enriched with high molecular mass as the library size increases.