Accessibility QTL

Per-variant predictions on two chromatin-accessibility QTL benchmarks — caQTL (ATAC; DeGorter et al. 2023) and dsQTL (DNase-I sensitivity; Degner et al. 2012) — scored with the supervised official metrics (#311). Two metrics per model: causality auPRC (significant-QTL vs control, ranking variants by |score|) and direction Pearson (signed correlation of the variant-effect score with the measured study effect, over the positives only). Higher = better. Numbers are the train (odd chroms) dev split; the even-chrom test set is held out for final eval.

Dataset

Leaderboard

Cells are the metric value; color ranks models within each column (own 0→max scale — the two metrics are not on a shared scale). Click a header to sort; hover for ± SE.