Benchmarking of deep learning-facilitated phosphopeptide identification on the synthetic dataset.A, three phosphopeptide identification methods were benchmarked. B, the number of correctly localized PSMs at different levels of PSM FLR are shown for the three methods, respectively, together with Methods 1 and 4. The numbers of correctly localized PSMs at 1% FLR and the percent increase compared with Method 1 (phosphoRS) are indicated. C, upSet plot comparing the PSM identifications from all seven methods. The number of correctly localized PSMs by Method 7 is marked as red. D, the FLRs at different site probability cutoffs are shown. The FLRs at 0.75 site probability are indicated.