Skip to content

The principle behind FastaLake#

De novo sequencing predicts peptide sequences directly from spectra, without matching them to a protein reference. FastaLake uses those predictions to select complete proteins from a broad reference catalogue, then searches the measured spectra with Sage. This keeps the search space manageable and preserves a record of how each candidate entered it. Because the selected database contains whole proteins, Sage can identify additional peptides beyond the predictions used for selection. Recovery still depends on which sequences the reference contains.

How the workflow connects#

flowchart TD
    R["Protein reservoir<br/>Complete sequences with source provenance"]
    P["De novo predictions<br/>One set per acquisition"]
    E["Study-supported sequence space<br/>Exact lookup of retained queries pooled across the study"]
    C["Acquisition-specific candidate set<br/>Exact lookup of this acquisition's retained queries"]
    D["Selected acquisition-specific database<br/>Selection parsimony retains complete proteins"]
    F["Final target FASTA"]
    M["Optional matched DNA / RNA evidence<br/>Add exact sequences after selection"]
    S["Sage database search<br/>Original spectra against the final FASTA"]
    A["Original spectra<br/>One mzML per acquisition"]
    G["Study-level protein groups<br/>Accepted peptides assigned to fixed representatives"]
    Q["Quantification and QC<br/>Sum or directLFQ; coverage, missingness and eligible PCA"]
    R --> E
    P --> E
    E --> C
    P --> C
    C --> D
    D --> F
    M -.-> F
    F --> S
    A --> S
    S --> G
    G --> Q
    classDef optional fill:#fff4de,stroke:#9b6b14,color:#302510
    classDef main fill:#edf6f5,stroke:#35766c,color:#173934
    class R,E,C,D,F,S,G,Q main
    class M optional
  • Evidence selection first reduces the reference using the study's predictions, then narrows candidates for each acquisition. Reference pieces merge before inference; each piece uses the same prediction-ranking rules.
  • Razor selection assigns mapped peptides using the declared deterministic rule and retains proteins receiving an assignment.
  • Sage searching evaluates the final candidates against measured spectra with explicit digestion, modification and confidence settings.
  • Study reporting connects accepted peptide evidence to fixed reporting groups, quantities and QC. The exported ledgers retain local protein memberships.

The algorithm guide specifies thresholds and tie rules; the function map links the code and tests behind each stage.

Names used in the results#

The glossary defines each sequence space and connects it to the actual file or command name. The illustrated workflow shows how the spaces change, step by step.

How multi-omics adds evidence#

Matched DNA and RNA select genes independently within a specimen-qualified reference. Their exact protein sequences are added after razor and before Sage, preserving every de novo baseline protein and header. Optional metabolite links provide another explicit selection route. These inputs nominate candidates; they do not themselves establish protein detection, reaction flux or organism of origin. The molecular tutorial explains preparation and provenance.

Interpretation and practical scope#

Shared peptides can leave biological origin ambiguous. A representative accession is therefore not automatically a unique protein or species identification. Likewise, a blank quantity is not evidence of absence. Adding candidates may gain some search identifications and lose others. Retaining baseline sequences does not guarantee identical search acceptance.

Database selection uses study evidence, so a nominal search q-value does not alone calibrate the whole adaptive procedure. The scientific evidence record is kept with the private review record (not part of this repository).

Clusters remain valuable for large references, broad searches and sustained cohort processing. Chunking reduces extraction memory but does not bound every stage. Use the measured resource guide to choose an execution environment, and the Docker tutorial to run the same workflow locally.