Peptide-spectrum match (PSM) rescoring is a powerful enhancement of standard searching that boosts the spectrum annotation performance. Recently we fine-tuned the deep learning-based fragment ion intensity prediction model, Prosit, on a tryptic and non-tryptic ground-truth dataset acquired on timsTOF instruments, expanding the utility of Prosit to this data type. Implementing the new model into the PSM rescoring tool Oktoberfest resulted in an up to 3-fold improvement in the identification of immunopeptides and an improved identification rate of samples digested with AspN, GluC, and trypsin, demonstrating a broader biological and biomedical application.

The high sensitivity of timsTOF instruments makes them ideal for studying samples containing low-abundance peptides. The parallel accumulation-serial fragmentation (PASEF) mode enables repeated re-targeting of low-abundance precursors. These individual scans of the same precursor are then summed to improve MS/MS spectrum quality. However, this relies on search engine-specific implementations of scan summing functionality.

Until now, Oktoberfest for timsTOF data was limited to search engine outputs from MaxQuant, as the timsTOF Prosit 2023 model relies on MaxQuant for summing the MS/MS scans. Here, we present an extension of Oktoberfest to support MSFragger, a faster and well-established search engine, thereby significantly increasing its usability.

Before implementing MSFragger into Oktoberfest, we will evaluate whether the timsTOF Prosit 2023 model can be used on spectra resulting from a different scan summing method. In addition, we will compare the results from MSFragger followed by Oktoberfest with those previously obtained with MaxQuant and Oktoberfest, as well as results from MSFragger and MSBooster. We anticipate that this improvement will especially benefit researchers investigating low-abundance peptides or working with large datasets.