Based on two deep, raw-data–grounded highlights you provided—high-throughput single-cell proteomics with RTLS + TMTpro 32-plex () and a full-isolation-window DDA search strategy that targets co-fragmented-peptide identification ()—the author’s work appears strongly oriented toward (i) throughput/depth engineering in proteomics and (ii) search/identification accuracy improvements in complex MS data.
Date context: April 29, 2026. Evidence used below is limited to the two BGPT-provided raw-data–grounded paper excerpts you supplied (plus their stated DOI-level identifiers). Where the provided record is incomplete, I mark uncertainty explicitly.
What’s most defensible from the provided evidence
Proteomics engineering for throughput at single-cell scale is strongly supported by the provided RTLS + TMTpro 32-plex + dual-column LC + nanoPOTS description, including explicit per-cell and imputed protein quantification counts in a 2,130-cell PBMC dataset ().
MS identification-method innovation is supported by the full isolation window search concept in MSFragger-DDA+, with the key methodological claim being improved sensitivity via detection of co-fragmented peptides followed by targeted extraction and rescoring ().
Explicit consideration of limitations appears in the RTLS single-cell work, including single-donor generalizability, library-matching biases, assumptions in proteomics–scRNA-seq integration, and technical sources of low-protein clusters ().
1) Visualize the single-cell proteomics throughput/depth claims (from provided raw excerpt)
Numbers below are exactly those present in your provided excerpt for the RTLS scProteomics PBMC paper (). Because your excerpt includes potentially conflicting median labels (e.g., “median peptide identifications per cell: 752” and separately “median peptides per cell: 2164”), I treat this visualization strictly as “as-stated in excerpt” and call out the ambiguity later.
2) Visualize throughput framing (cells/day and dataset size)
These are excerpt-derived quantities. Without the full paper text here, I cannot verify additional details (e.g., whether “~660 cells/day” includes all operational overheads) beyond what your excerpt states ().
3) Algorithmic innovation: what MSFragger-DDA+ is claiming (and what to be skeptical about)
The provided record indicates that MSFragger-DDA+ performs a full isolation window search, then performs targeted precursor signal extraction and greedy rescoring to improve peptide identification sensitivity by detecting co-fragmented peptides within wide-window acquisition ().
Skeptical reading: sensitivity gains can depend on acquisition conditions (isolation width, co-fragmentation rates), dataset composition, and evaluation set bias. Your excerpt explicitly notes “reliance on specific datasets” and the need for further validation in diverse proteomic contexts ().
4) Critical synthesis: scientific strength vs. blindspots
Strength: engineering-to-measurement loop. The RTLS single-cell work is not just descriptive; it reports concrete throughput and depth numbers, and it frames falsification targets (e.g., no improvement vs standard MS2, loss of cell-type resolution, or issues from bridging/label matching) in the provided excerpt ().
Strength: method claims align with MS realities. MSFragger-DDA+ explicitly addresses co-fragmentation by searching the full isolation window and rescoring after signal extraction—an approach plausibly consistent with how co-fragmented spectra can confound standard DDA interpretations ().
Blindspot: evaluation-set and generalizability. Both excerpts emphasize—explicitly in RTLS (single-donor PBMC) and implicitly in DDA+ (dataset dependence)—that performance may not transfer across donors, tissues, instruments, or acquisition settings (;).
Blindspot: proteomics–transcriptomics mapping assumptions. The RTLS excerpt states multimodal reference mapping with an scRNA-seq atlas and then clustering/mapping cell types. That pipeline can propagate reference atlas biases; if protein–RNA concordance shifts across conditions, mapping might mislabel ambiguous states ().
Data ambiguity to resolve (from excerpt only): your RTLS excerpt contains two different “median peptides per cell” values (one says “median peptide identifications per cell: 752” and another says “median peptides per cell: 2164”). I visualized the “median peptides per cell: 2164” and “median proteins per cell: 752” entries, but I cannot reconcile this inconsistency without the full paper tables/figure captions ().
The RTLS scProteomics excerpt explicitly reports financial interests involving MSFragger and IonQuant licensing, and F. Yu receiving paid consultancy and financial interest through licensing of MSFragger and IonQuant to commercial entities (). The MSFragger-DDA+ excerpt also reports royalties/financial interest tied to MSFragger, IonQuant, and diaTracer licenses ().
Why this matters scientifically: COI does not invalidate results, but it increases the need for skeptical evaluation of whether algorithmic improvements are tested broadly, transparently, and with fair baselines. The RTLS excerpt partly addresses this by stating key limitations and potential falsification targets ().
6) How to best verify/strengthen this author review (next step)
To make this review maximally truth-oriented, you should check (i) whether the DDA+ sensitivity gains hold on truly out-of-distribution acquisition settings, (ii) whether RTLS scProteomics replicates across multiple donors/tissues, and (iii) whether proteomics–scRNA label transfer shows systematic disagreement in known perturbation contexts.
It will extract the excerpted scProteomics counts, reconcile per-metric ambiguities, and generate labeled plots comparing peptides, proteins, and imputed proteins for rapid evidence review of 10.64898/2026.02.01.703169.
Get emailed when your analysis is done!
We'll email you the results when your analysis is finished.
Hypothesis Graveyard
A “universal improvement regardless of acquisition regime” hypothesis for DDA+ is likely wrong; the excerpt itself flags dataset/complexity dependence and the need for broader validation.
A “single-donor results generalize automatically to diverse immune states” hypothesis is unlikely; the RTLS excerpt explicitly notes single-donor generalizability limits.
Quick Explanation
Long Explanation
Top Data Sources
ExportMCP
Keep Exploring
Analysis Wizard
Hypothesis Graveyard
Potential Experiments
Discussion
Fuel Your Discoveries
Ask a Follow-Up
Key Insight