Add CADRE model with tests, docs, examples, and benchmark

[chadwickdonald]

Contributor

Chadwick Bidwell
bidwell3@illinois.edu

Contribution Type

Model

Original Paper

Predicting Drug Sensitivity of Cancer Cell Lines via
Collaborative Filtering with Contextual Attention
https://proceedings.mlr.press/v126/tao20a/tao20a.pdf

Summary

This PR adds a PyHealth-native implementation of CADRE for multilabel drug-response prediction.

Features include:

• gene embedding encoder
• optional contextual attention
• drug-specific decoder
• multilabel outputs

Files to Review

• pyhealth/models/cadre.py
• pyhealth/models/init.py
• tests/core/test_cadre.py
• examples/sample_multilabel_cadre.py
• examples/benchmark_cadre_vs_mlp.py
• docs/api/models/pyhealth.models.cadre.rst
• docs/api/models.rst

Validation

• unit tests pass
• example scripts run successfully
• benchmark compares CADRE variants vs MLP baseline

Notes

This implementation is inspired by the original CADRE paper/repository