Search papers, labs, and topics across Lattice.
Amazon ScienceDepartment of Computer ScienceDepartment of Medical Biochemistry and BiophysicsDepartment of Quantitative and ComputationalUSCJun 9, 2026arXiv:2606.11382The GLACIER model integrates multiple data modalities鈥攎olecular graphs, SMILES strings, and physicochemical descriptors鈥攖hrough a student-teacher framework to enhance molecular property prediction. By pretraining specialized student encoders and employing a Finsler geometry-aware module for fusion, GLACIER effectively distills knowledge from larger teacher models using contrastive learning. This approach not only improves predictive performance but also significantly reduces computational demands, making it scalable for real-world applications in molecular discovery.
GLACIER achieves high predictive accuracy while drastically cutting down the computational costs typically associated with multimodal molecular property prediction.
Deep learning models facilitate the discovery of molecules with tailored properties among billions of candidate compounds. However, the computational burden to develop and deploy state-of-the-art models continuously increases, limiting their scalability. Most large-scale models are unimodal in nature and overlook the potential to leverage complementary molecular data modalities. To address these shortcomings, this paper introduces the Graph-Language Alignment for Chemical Inference and Exploration using Representations (GLACIER) model, a student-teacher framework that integrates molecular graphs, SMILES strings, and physicochemical descriptors to learn rich molecular embeddings. Our framework consists of three stages: (1) we pretrain three student encoders on 100,000 drug-like molecules: a message-passing neural network for molecular graphs, a transformer-based encoder for SMILES strings, and a multilayer perceptron for physicochemical descriptors, (2) we fuse these student modalities using a novel Finsler geometry-aware module, and (3) distill complementary knowledge from large teacher models, including MiniMol and MolFormer, into a single lightweight model via contrastive learning. We demonstrate that GLACIER is a robust framework that delivers high predictive performance and computational efficiency in complex molecular property prediction tasks. Our code is publicly available at https://github.com/eemokey/glacier.
