Identify more peptides from your single-cell data

Single-cell and low-input proteomics leaves little material to work with, so every spectrum counts. This application note shows how Tesorai Search performs on two public single-cell datasets compared with FragPipe at 1% FDR. Match-between-runs was disabled for both tools, so the gains reflect identifications from the acquired spectra themselves.

Download the application note (PDF)

Make more of limited starting material

Tesorai Search scores peptide-spectrum matches with a pretrained model and controls error with target-decoy false discovery rate estimation. On NanoPOTS and DISCO single-cell data, it identified more peptides and more MS2 spectra than FragPipe from the same runs.

Benchmark results

NanoPOTS (Williams et al., MCF10A)

Unique peptides, Tesorai vs. FragPipe:

• 1 cell: 2,472 vs. 1,910 (+29%)
• 3 cells: 4,301 vs. 3,611 (+19%)
• 10 cells: 6,614 vs. 5,620 (+18%)
• 50 cells: 13,030 vs. 10,720 (+22%)

Counts are pooled across runs within each input condition, not per-cell averages. One cell: 10 runs; 3, 10 and 50 cells: 4 runs each.

DISCO (Lamanna et al.)

MS2 identification rate, Tesorai vs. FragPipe:

• 1 cell: 15.5% vs. 11.3% (+37%), with 6,273 vs. 5,713 unique peptides
• 5 cells: 23.7% vs. 17.4% (+36%), with 10,729 vs. 9,598 unique peptides

Three runs per condition. Tesorai counts exclude decoys and non-human contaminants.

Scale your analysis without managing compute

Run single-cell and low-input database searches through your browser. Tesorai handles processing in the cloud, with no search software to install and no local compute infrastructure to maintain.

1. Upload your data. Add mass spectrometry files and select a sequence database.
2. Configure and run. Set search parameters and run the search in the cloud.
3. Review your results. Explore peptide and spectrum identifications.

Methods

Benchmarks measure peptide and spectrum identification sensitivity, not protein counts. NanoPOTS: MCF10A, MassIVE MSV000085230. DISCO: PRIDE PXD019958, Thermo Q Exactive HF-X. Both datasets were searched against a human Swiss-Prot database with default Tesorai settings; FragPipe v22.0 used strict trypsin specificity and peptide lengths of 7–63 amino acids.

References

1. Burq M, et al. Tesorai Search. J Mol Biol (2026), Fig. 2 and supplementary methods and tables. doi:10.1016/j.jmb.2026.169940
2. Lamanna J, et al. Digital microfluidic isolation of single cells for -Omics. Nat Commun 11, 5632 (2020). doi:10.1038/s41467-020-19394-5
3. Williams SM, et al. Automated coupling of nanodroplet sample preparation with LC-MS for high-throughput single-cell proteomics. Anal Chem 92, 10588–10596 (2020). doi:10.1021/acs.analchem.0c01551
4. Tesorai Search benchmark analysis of DISCO dataset PXD019958. Target identifications exclude decoys and non-human contaminants; match-between-runs disabled for both tools.