Platform

Architecture and relationship to TDC

PyTDC extends Therapeutics Data Commons with an API-first dataset layer, a model server for retrieval and inference, and single-cell therapeutics tasks. TDC remains the dataset commons; PyTDC is the training, evaluation, and inference platform built on top of it.

Model server

Diagram of the PyTDC model server connecting data sources, model weights, and evaluation modules
Figure from the PyTDC paper: the model server retrieves weights, exposes inference endpoints, and feeds embeddings into task-specific benchmarks.

Workflow

Diagram of a PyTDC research workflow from data loading through inference
Paper figure: a compact workflow from dataset loading to model inference.

Software