alterlab-pytdc

You can load AI-ready drug discovery datasets by using the provided Python tdc package utilities to access and prepare therapeutics data for reproducible model evaluation across various prediction tasks.

Therapeutics machine learning tasks support scaffold, random, cold, and temporal splits to ensure robust model evaluation across different chemical structures and time-based data distributions.

Yes, you can benchmark molecular generation and retrosynthesis models using the provided generation task utilities, which include standard datasets and evaluation functions for reproducible experiments.

Yes, numpy and pandas are required dependencies to process ADME and toxicity datasets, as they provide the foundational data manipulation structures for the benchmarking workflows.

The best way to evaluate drug-target interaction prediction models reproducibly is by using standardized benchmark groups with predefined data splits and built-in evaluation utilities for consistent results.

Yes, PyTDC supports multi-instance prediction tasks across the therapeutics pipeline, covering areas like DDI and PPI, with standardized data loading and evaluation utilities for comprehensive benchmarking.