Cross-Study Analysis

Cross-Study Analysis (CSA) looks at the ability of models to generalize across different datasets. For each model, the CSA workflow generates an N-by-N table of prediction performance scores (N = number of datasets). Scores on the diagonal indicate performance when trained and tested on different splits of the same dataset, whereas off-diagonal scores indicate performance when trained and tested on different datasets

Workflow

Metrics

References

1. A. Partin et al. “Systematic evaluation and comparison of drug response prediction models: a case study of prediction generalization across cell lines datasets”, AACR, 2023 2. F. Xia et al. “A cross-study analysis of drug response prediction in cancer cell lines”, Briefings in Bioinformatics, 2022