PyTDC: A multimodal machine learning training, evaluation, and inference platform for biomedical foundation models

Alejandro Velez-Arce, Marinka Zitnik

TL;DR: PyTDC is a machine learning platform for training, evaluation, and inference on multimodal biomedical AI models, integrating single-cell data, benchmarks, and a model server to enable context-aware model development for therapeutic discovery.

Abstract: Existing biomedical benchmarks do not provide end-to-end infrastructure for training, evaluation, and inference of models that integrate multimodal biological data and a broad range of machine learning tasks in therapeutics. We present PyTDC, an open-source machine-learning platform providing streamlined training, evaluation, and inference software for multimodal biological AI models. PyTDC unifies distributed, heterogeneous, continuously updated data sources and model weights and standardizes benchmarking and inference endpoints. This paper discusses the components of PyTDC's architecture and, to our knowledge, the first-of-its-kind case study on the introduced single-cell drug-target nomination ML task. We find state-of-the-art methods in graph representation learning and domain-specific methods from graph theory perform poorly on this task. Though we find a context-aware geometric deep learning method that outperforms the evaluated SoTA and domain-specific baseline methods, the model is unable to generalize to unseen cell types or incorporate additional modalities, highlighting PyTDC's capacity to facilitate an exciting avenue of research developing multimodal, context-aware, foundation models for open problems in biomedical AI.

Lay Summary: Biomedical research for new treatments involves analyzing complex data from sources like genes, proteins, and clinical trials. However, combining this diverse data and using it with machine learning is difficult because existing tools don’t fully support the process or keep up with new information. Our team created PyTDC, a platform that brings together these varied data sources and offers simple tools for training, testing, and applying machine learning models. PyTDC keeps data current with an innovative design and provides a model server to access and compare cutting-edge models easily. This helps researchers and developers study diseases and treatments more effectively. By simplifying access to data and models, PyTDC could speed up the discovery of new therapies and improve our understanding of diseases, making a real difference in healthcare.

Application-Driven Machine Learning: This submission is on Application-Driven Machine Learning.

Link To Code: https://github.com/apliko-xyz/PyTDC

Primary Area: Applications->Health / Medicine

Keywords: biological chemistry, biomedical AI, computational biology, biomedical informatics, machine learning platform, machine learning framework, AI for drug discovery

Submission Number: 12343