MS2-Based Evaluation of Precursor-Specific Fragmentation Conditions for Improving TMT Reporter Signal in MS3-Based Quantification

Tandem mass tag (TMT)-based quantitative proteomics is a leading approach for multiplexed protein quantification, with synchronous precursor selection (SPS)-MS3 as the state-of-the-art strategy to mitigate ratio compression caused by precursor coisolation. Because TMT reporter signal in SPS-MS3 is directly proportional to the abundance of SPS ions selected from the preceding MS2 scan, maximizing SPS ion production should improve quantification. However, current data acquisition strategies apply a universal MS2 fragmentation setting to all peptide precursors, even though peptides possess distinct physicochemical properties that govern SPS ion production and, consequently, quantitative TMT reporter ion signal. While various fragmentation strategies have been explored to maximize peptide identification in label-free MS2 contexts, how to maximize quantitative signal in SPS-MS3-TMT remains underexplored. Inspired by recent advances in intelligent, real-time data acquisition, we hypothesize that adaptive fragmentation tuned per precursor could improve SPS-MS3-based TMT quantification. Here, we present a systematic evaluation using a TMTpro-labeled human data set of 51,064 unique tryptic peptide precursors, each acquired under ten fragmentation conditions (CID and HCD across a range of normalized collision energies). From each MS2 spectrum, we then computationally selected up to ten fragment ions ranked by intensity as pseudo-SPS (p-SPS) ions and used their summed intensity as a proxy for reporter ion signal. We find that CID NCE25, rather than the widely used CID NCE34, maintains identification efficiency and serves as the single best default setting, optimal for p-SPS ion production in more precursors (24%) than any other individual setting. In addition, per-precursor fragmentation optimization increases p-SPS ion signal by ≥25% for 36% of precursors and, more strikingly, ≥50% for 13% of precursors. To enable practical implementation, we developed FragWise, an XGBoost multiclass classifier that predicts the optimal fragmentation setting from precursor features, achieving 91.6% of the maximum achievable p-SPS ion signal. Together, these findings demonstrate the potential of differential fragmentation and lay the groundwork for real-time adaptive fragmentation to maximize quantitative information in TMT-based proteomics.