Table 7 AUROC scores for the expression quantitative trait loci prediction (eQTLP) task across different cell types

From: DNALONGBENCH: a benchmark suite for long-range DNA prediction tasks

Models

Expression Quantitative Trait Loci Prediction

 

CCF

WB

Thyroid

SNSES

SSELL

MS

NT

AT

AS

Avg

Expert Model

0.639

0.689

0.612

0.710

0.700

0.621

0.683

0.741

0.736

0.681

CNN

0.547

0.577

0.487

0.499

0.499

0.502

0.516

0.576

0.551

0.528

HyenaDNA

0.584

0.512

0.529

0.471

0.544

0.487

0.511

0.479

0.513

0.514

Caduceus-Ph

0.597

0.594

0.527

0.586

0.574

0.538

0.588

0.547

0.541

0.565

Caduceus-PS

0.549

0.542

0.547

0.529

0.541

0.523

0.552

0.536

0.519

0.537

  1. The cell types are abbreviated as follows: CCF for Cells_Cultured_fibroblasts, WB for Whole_Blood, SNSES for Skin_Not_Sun_Exposed_Suprapubic, SSELL for Skin_Sun_Exposed_Lower_leg, MS for Muscle Skeletal, NT for Nerve Tibial, AT for Artery_Tibial, and AS for Adipose_Subcutaneous. Notably, Expert Model achieves the best performance across all cell types. The highest scores are highlighted in bold. “Avg” represents the average score across all cell types.
Back to article page