FastaLake default pipeline Study-wide exact protein lookup; per-acquisition candidate retrieval, selection parsimony and Sage search; shared study groups, summed intensities and quality control. FastaLake Protein selection, search and study quantification Rust selection core FOR THE STUDY Protein reservoir FASTA · complete proteins De novo predictions CSV Sequences predicted from spectra Within each acquisition, retain positive-score predictions of 9–50 residues after the complexity filter, using the top-30% score threshold and retaining ties. Pool distinct I/L-equivalent queries and retrieve complete proteins by exact containment with Aho–Corasick. Find study-supported proteins Retrieve complete proteins by exact peptide lookup. Isoleucine and leucine are treated as equivalent. 9–50 residues · top 30% per run; score ties retained FOR EACH ACQUISITION All eligible predictions · CSV Measured spectra · mzML Use the same lookup filter on each acquisition’s predictions to retrieve candidates from the study-supported protein space. Lookup uses the entire predicted sequence, with isoleucine and leucine equivalent and no enzyme constraint. Find acquisition candidates Find proteins containing this run’s filtered predictions. Exact lookup · Rust Selection parsimony uses all eligible predictions mapping to the candidates, including those below the lookup score threshold. Fixed unique- and total-evidence counts rank compatible proteins; accession hashes resolve ties. Retain complete proteins receiving evidence. This operation is separate from post-search study grouping. Select the search database Keep full proteins that explain the mapped prediction evidence. Selection parsimony · Rust Sage 0.14.6 searches one acquisition at a time against its selected FASTA. The default is closed tryptic searching over peptides of 5–50 residues, with generated decoys and native label-free intensities. Target peptides pass at peptide q ≤ 0.01 before study grouping. The lookup minimum of nine residues is not a final-identification length filter. Search measured spectra Match spectra to the selected protein database. Sage 0.14.6 Peptide error cutoff: q ≤ 0.01 Search-summary tables ACROSS THE STUDY A deterministic greedy cover of the accepted peptide–protein graph defines a single study dictionary, restricted to the FASTAs actually searched. Each canonical peptide is assigned to the first selected representative explaining it. Sum positive finite LFQ feature intensities only where that peptide was accepted in the same acquisition, preserving missing values and counting each admitted feature once. Study groups and intensities Assign accepted peptides to shared study groups; sum their intensities within each run. Group × acquisition matrix · TSV AlphaPeptTools 0.4.0 reports detection, intensity and missingness without extra normalization or imputation. PCA runs on complete varying groups when at least three nonempty acquisitions and two eligible groups are available; otherwise the output records why it was skipped. Quality control Review detection and intensity. PCA summarizes variation, when eligible. AlphaPeptTools · reports Acquisition = one MS run Default workflow · code e072ca2
FastaLake default pipeline Study-wide exact protein lookup; per-acquisition candidate retrieval, selection parsimony and Sage search; shared study groups, summed intensities and quality control. FastaLake Protein selection, search and study quantification Rust selection core FOR THE STUDY Protein reservoir FASTA · complete proteins De novo predictions CSV Sequences predicted from spectra Within each acquisition, retain positive-score predictions of 9–50 residues after the complexity filter, using the top-30% score threshold and retaining ties. Pool distinct I/L-equivalent queries and retrieve complete proteins by exact containment with Aho–Corasick. Find study-supported proteins Retrieve complete proteins by exact peptide lookup. Isoleucine and leucine are treated as equivalent. 9–50 residues · top 30% per run; score ties retained FOR EACH ACQUISITION All eligible predictions · CSV Measured spectra · mzML Use the same lookup filter on each acquisition’s predictions to retrieve candidates from the study-supported protein space. Lookup uses the entire predicted sequence, with isoleucine and leucine equivalent and no enzyme constraint. Find acquisition candidates Find proteins containing this run’s filtered predictions. Exact lookup · Rust Selection parsimony uses all eligible predictions mapping to the candidates, including those below the lookup score threshold. Fixed unique- and total-evidence counts rank compatible proteins; accession hashes resolve ties. Retain complete proteins receiving evidence. This operation is separate from post-search study grouping. Select the search database Keep full proteins that explain the mapped prediction evidence. Selection parsimony · Rust Sage 0.14.6 searches one acquisition at a time against its selected FASTA. The default is closed tryptic searching over peptides of 5–50 residues, with generated decoys and native label-free intensities. Target peptides pass at peptide q ≤ 0.01 before study grouping. The lookup minimum of nine residues is not a final-identification length filter. Search measured spectra Match spectra to the selected protein database. Sage 0.14.6 Peptide error cutoff: q ≤ 0.01 Search-summary tables ACROSS THE STUDY A deterministic greedy cover of the accepted peptide–protein graph defines a single study dictionary, restricted to the FASTAs actually searched. Each canonical peptide is assigned to the first selected representative explaining it. Sum positive finite LFQ feature intensities only where that peptide was accepted in the same acquisition, preserving missing values and counting each admitted feature once. Study groups and intensities Assign accepted peptides to shared study groups; sum their intensities within each run. Group × acquisition matrix · TSV AlphaPeptTools 0.4.0 reports detection, intensity and missingness without extra normalization or imputation. PCA runs on complete varying groups when at least three nonempty acquisitions and two eligible groups are available; otherwise the output records why it was skipped. Quality control Review detection and intensity. PCA summarizes variation, when eligible. AlphaPeptTools · reports Acquisition = one MS run Default workflow · code e072ca2

A closer look at each stage

Find study-supported proteins

Within each acquisition, retain positive-score predictions of 9–50 residues after the complexity filter, using the top-30% score threshold and retaining ties. Pool distinct I/L-equivalent queries and retrieve complete proteins by exact containment with Aho–Corasick.

Find acquisition candidates

Use the same lookup filter on each acquisition’s predictions to retrieve candidates from the study-supported protein space. Lookup uses the entire predicted sequence, with isoleucine and leucine equivalent and no enzyme constraint.

Select the search database

Selection parsimony uses all eligible predictions mapping to the candidates, including those below the lookup score threshold. Fixed unique- and total-evidence counts rank compatible proteins; accession hashes resolve ties. Retain complete proteins receiving evidence. This operation is separate from post-search study grouping.

Search measured spectra

Sage 0.14.6 searches one acquisition at a time against its selected FASTA. The default is closed tryptic searching over peptides of 5–50 residues, with generated decoys and native label-free intensities. Target peptides pass at peptide q ≤ 0.01 before study grouping. The lookup minimum of nine residues is not a final-identification length filter.

Study groups and intensities

A deterministic greedy cover of the accepted peptide–protein graph defines a single study dictionary, restricted to the FASTAs actually searched. Each canonical peptide is assigned to the first selected representative explaining it. Sum positive finite LFQ feature intensities only where that peptide was accepted in the same acquisition, preserving missing values and counting each admitted feature once.

Quality control

AlphaPeptTools 0.4.0 reports detection, intensity and missingness without extra normalization or imputation. PCA runs on complete varying groups when at least three nonempty acquisitions and two eligible groups are available; otherwise the output records why it was skipped.

Inputs and scope

The workflow starts with a prepared protein reservoir, de novo prediction CSVs and one mzML file per acquisition. Raw-file conversion and de novo model inference are upstream. Selection and search run separately for each acquisition; the study dictionary is built afterward from the accepted peptides.

How to read the outputs

Peptide q ≤ 0.01 is the default Sage search-acceptance cutoff. It is not a measured error rate for the whole workflow. Study groups provide consistent reporting labels; a representative does not establish the unique protein or organism of origin. The additional search-summary tables use the search-reported member sets and remain separate from the primary study-group matrix.

Optional extensions

Summed intensities are the default. directLFQ 0.3.3 can be selected explicitly. Specimen-matched molecular evidence can add protein candidates after selection; optional eggNOG annotation describes the reporting representatives. These extensions are outside the default flow shown above.

Code and reproducibility

The figure follows MannLabs/fasta-lake at the recorded source commit. Both SVG themes and this page are generated from the same stage data. Fonts are embedded; the figure makes no external requests.

View the source workflow