dnoise: Fast Native Data Reduction for Bruker timsTOF

The central claim — that default MS1-only denoising removes a large fraction of native timsTOF frame data (35–53% of the frame binary) while leaving the tested proteomics results essentially unchanged — is supported by the evidence presented. For ddaPASEF it is close to mechanically guaranteed (the searched MS/MS spectra are untouched, and identification counts are identical), and the quantitative validation on a defined three-species mixture with known ratios, replicated six times per condition across two gradients and two acquisition modes, is a disciplined test of the part of the claim that could have failed. Runtime and memory claims are descriptive measurements on stated hardware and are correctly reported as such. The software is released under MIT with a versioned Zenodo archive, the raw data are public, and the round-trip fidelity check (point-for-point identity of an unfiltered pass) is exactly the control a reader wants to see for a tool that rewrites a proprietary binary. What lowers readiness is not the validity of the core result but the calibration of its presentation and several reporting gaps. The two geometric streak-filter parameters were selected on Condition A of the 15-minute ddaPASEF gradient, which then reappears inside the pooled headline result; the authors disclose this in Section 2.2 but the Abstract and Conclusion present a single pooled figure across all four arms without the qualifier. Related, the phrase "LFQ accuracy was preserved in both modes" implies symmetric evidentiary strength, whereas the diaPASEF case rests on a metric (fragment-based MaxLFQ) that MS1-only filtering does not directly touch, with the one MS1-sensitive check (Table S12) reported without the bootstrap treatment applied elsewhere. The DDA/DIA reduction differential is confounded by on-instrument denoising, which the manuscript warns against misreading but does not bound. Finally, the compliance audit identifies genuine gaps: search FDR thresholds are never stated, the analysis and figure-generation scripts are described but not deposited, the optional-mode `msms_*` parameters and centroider algorithms are underspecified, and two references have anomalous dates or DOIs. Every one of these is fixable in text, tables, and deposited files. None requires a new acquisition, a new search, or a reanalysis whose outcome could overturn a conclusion. Where I would otherwise have asked for new work — a second sample type, a fair-stringency fragment control, a mechanistic test of the decoy-ratio inversion — the correct remedy at this venue is to scale the claim to the evidence rather than to demand the experiment, and I have routed those to suggestions and to a requalification requirement accordingly. That places this at minor revision. The score reflects a sound, useful, honestly reported tool paper whose current text overreaches its four benchmark arms in a few specific sentences and whose reproducibility record has repairable holes.

Readiness measures the current manuscript. Novelty, significance, and usefulness are shown separately. The recommendation follows the work required for publication, not a score range.