Personal transcriptome variation is poorly explained by current genomic deep learning models.

We find a small trend towards larger cross-individual correlations for genes with smaller distance between the most statistically significant eQTL and the TSS (Fig. 2c and Extended Data Fig. 8 ), which aligns with previous findings that current sequence-to-expression models capture gene expression determinants in promoters more accurately than distal enhancers 16 .

also in 150 other papers