Geometric Transformer for End-to-End Molecule Properties Prediction
This work addresses the problem of predicting molecule properties for researchers in chemistry and drug discovery, representing an incremental improvement by adapting existing Transformer methods to molecular data.
The authors tackled the challenge of applying Transformers to non-sequential molecular data by introducing a geometry-aware Transformer architecture that captures molecular geometry through modified positional encoding and gated self-attention, outperforming state-of-the-art methods without relying on domain-specific knowledge.
Transformers have become methods of choice in many applications thanks to their ability to represent complex interactions between elements. However, extending the Transformer architecture to non-sequential data such as molecules and enabling its training on small datasets remains a challenge. In this work, we introduce a Transformer-based architecture for molecule property prediction, which is able to capture the geometry of the molecule. We modify the classical positional encoder by an initial encoding of the molecule geometry, as well as a learned gated self-attention mechanism. We further suggest an augmentation scheme for molecular data capable of avoiding the overfitting induced by the overparameterized architecture. The proposed framework outperforms the state-of-the-art methods while being based on pure machine learning solely, i.e. the method does not incorporate domain knowledge from quantum chemistry and does not use extended geometric inputs besides the pairwise atomic distances.