Skip to main content

Table 4 The performance of protein domain prediction with out-of-distribution examples using ProDOMA with Outlier Exposure (OE), and HMMER on the real PacBio and Nanopore dataset

From: Improving protein domain classification for third-generation sequencing reads using deep learning

Proportion Method PacBio Nanopore
   Recall Precision F1-score Recall Precision F1-score
1% GPCR HMMER 0.1403 0.9574 0.2446 0.3902 0.9876 0.5593
  ProDOMA 0.4581 0.9872 0.6258 0.4727 0.9796 0.6376
5% GPCR HMMER 0.1537 0.9411 0.2642 0.4015 0.9754 0.5513
  ProDOMA 0.4712 0.9921 0.6389 0.4615 0.9872 0.6289
10% GPCR HMMER 0.1491 0.9513 0.2577 0.3988 0.9917 0.5688
  ProDOMA 0.4473 0.9842 0.6150 0.4886 0.9843 0.6529
50% GPCR HMMER 0.1494 0.9507 0.2583 0.3984 0.9825 0.5670
  ProDOMA 0.4479 0.9837 0.6154 0.4836 0.9731 0.6458