Graph transformer model improves molecular chirality identification
$\text{GSF-}χ$: Global Stereochemical Fields for Chiral Graph Transformers
Machine Learning
Summary
Molecules can have mirror-image forms called enantiomers that behave differently, especially in chiral environments like the human body. The authors introduced a new method called GSF-χ that helps computers better detect these subtle differences by considering global stereochemistry, rather than focusing on single atoms. Their method respects molecular symmetry and reflection properties to predict molecular behaviors more accurately. This model shows improved results on tasks involving molecular rotations and chiral signals compared to previous approaches.
What this means in practice
- •For pharmaceutical developers: Predict molecular properties related to drug chirality for better design and testing of enantiomer-specific effects.$Commercial implications: Enables development of drug candidates with improved chiral selectivity and safety profiles through enhanced molecular encoding.
- •For chemical simulation engineers: Improve simulations of chiral molecular interactions by embedding global stereochemical information into models.
Authors
Jiaqing Xie, Yuxin Wang, Xipeng Qiu
Abstract
Enantiomers share atoms, bonds, and pairwise distances yet can behave differently in chiral environments, so molecular encoders must respect atom relabelings and proper rotations without becoming blind to reflection. We introduce GSF-$χ$, a graph transformer in which stereogenic units modulate all pairwise interactions rather than single out one atom as special. Each central or axial stereogenic unit creates a reflection-even phase field over all atoms, a handedness pseudoscalar $χ$ sets the direction of a relative rotation on latent query--key blocks, giving a \textbf{Chiral-RoPE} that reflection inverts rather than leaves fixed. A $C_2$ projection separates mirror-even ECD peak counts and positions from mirror-odd peak signs. We prove the operator's even--odd decomposition and its annotation-inversion, permutation, and unit-order identities under explicit canonical-role conditions; property tests and a coordinate-reflection audit verify the laws end to end. GSF-$χ$ leads every central-ECD output and improves axial Rotation and Symbol by $12.6\%$ and $7.9\%$ over the strongest baseline. Equal-budget controls attribute the Rotation advantage to global signed support rather than parameter or edge count; the $C_2$ projection yields exact enantiomer-pair consistency at a small raw-accuracy cost under complete supervision and becomes predictive when mirror supervision is scarce.