Study runner: fasta-lake run

The runner takes an acquisition manifest and a reference lake through evidence extraction, per-acquisition selection, Sage searching, study grouping, quantification and QC, writing every receipt on the way. Your first study shows it in use; Prepare an acquisition manifest is the input contract. The option list below is generated from the code.

usage: fasta-lake run [-h] --manifest MANIFEST [--lake LAKE]
                      [--lake-headers LAKE_HEADERS] --out OUT [--sage SAGE]
                      [--threads THREADS] [--laptop]
                      [--memory-budget-gib MEMORY_BUDGET_GIB]
                      [--memory-fraction MEMORY_FRACTION]
                      [--cpu-fraction CPU_FRACTION]
                      [--reference-chunk-mib REFERENCE_CHUNK_MIB]
                      [--top-fraction TOP_FRACTION] [--min-length MIN_LENGTH]
                      [--max-length MAX_LENGTH]
                      [--digestion {tryptic,semi-tryptic,unspecific}]
                      [--search-mode {closed,open}]
                      [--search-min-length SEARCH_MIN_LENGTH]
                      [--search-max-length SEARCH_MAX_LENGTH] [--databases-only]
                      [--member-set-aggregate] [--skip-qc] [--validate-only]
                      [--quantification {sum,directlfq}]
                      [--molecular-reference {augment,shared}]
                      [--molecular-dna-percent MOLECULAR_DNA_PERCENT]
                      [--molecular-rna-percent MOLECULAR_RNA_PERCENT]
                      [--molecular-dna-absolute MOLECULAR_DNA_ABSOLUTE]
                      [--molecular-rna-absolute MOLECULAR_RNA_ABSOLUTE]
                      [--molecular-metabolites] [--background BACKGROUND]
                      [--annotation-settings ANNOTATION_SETTINGS]

Portable, explicit Rust workflow for a declared acquisition manifest. Inputs and prior
outputs are preserved. Chunked extraction reclaims its own consumed scratch pieces. A
run requires a new output directory; failed runs remain available for diagnosis.
Installed as ``fasta-lake run``; ``tools/run_manifest.py`` remains as a compatibility
wrapper.

options:
  -h, --help            show this help message and exit
  --manifest MANIFEST   Tab-separated acquisition manifest: sample, predictions, mzml
  --lake LAKE           Required reservoir in augment mode
  --lake-headers LAKE_HEADERS
                        Header sidecar written by lake_builder --output-headers; names
                        study groups by gene, protein and organism (otherwise the
                        searched FASTA headers are used)
  --out OUT             New output directory (must not exist)
  --sage SAGE           Explicit path to the official SAGE v0.14.7 Linux x86_64
                        executable, which reports 'sage 0.14.6' (otherwise PATH)
  --threads THREADS     Rust threads (default: 4, automatic in laptop mode)
  --laptop              Detect resources and enable bounded reference pieces
  --memory-budget-gib MEMORY_BUDGET_GIB
                        Laptop planning budget; use Docker/Slurm for a hard RAM cap
  --memory-fraction MEMORY_FRACTION
                        Laptop share of visible RAM (default: 0.35)
  --cpu-fraction CPU_FRACTION
                        Laptop share of visible CPU slots (default: 0.5)
  --reference-chunk-mib REFERENCE_CHUNK_MIB
                        Extract reference pieces of this target size, then merge
                        before razor (e.g. 256)
  --top-fraction TOP_FRACTION
                        Share of top-scoring de novo predictions used as evidence
                        (default: 0.30)
  --min-length MIN_LENGTH
                        Shortest de novo peptide used (default: 9)
  --max-length MAX_LENGTH
                        Longest de novo peptide used (default: 50)
  --digestion {tryptic,semi-tryptic,unspecific}
                        Search cleavage specificity; does not change de novo selection
  --search-mode {closed,open}
                        Closed +/-20 ppm or open [-500,+100] Da precursor search
  --search-min-length SEARCH_MIN_LENGTH
                        Shortest searched peptide (default: 5)
  --search-max-length SEARCH_MAX_LENGTH
                        Longest searched peptide (default: 50)
  --databases-only      Stop after writing per-acquisition databases; no search or
                        grouping
  --member-set-aggregate
                        Also write the legacy member-set aggregate/ matrices.
                        Diagnostic only: they count distinct protein member sets,
                        which exceed protein groups
  --skip-qc             Explicitly omit automatic QC/PCA for a core-only installation
  --validate-only       Check inputs, executables and destination, then exit without
                        running
  --quantification {sum,directlfq}
                        Study quantification after searching (default: sum); QC/PCA
                        follows automatically
  --molecular-reference {augment,shared}
                        Augment a reservoir-derived baseline, or select from each
                        shared gene reference
  --molecular-dna-percent MOLECULAR_DNA_PERCENT
                        Top percent of positive available DNA genes; zero disables
                        (default: 1.0)
  --molecular-rna-percent MOLECULAR_RNA_PERCENT
                        Top percent of positive available RNA genes; zero disables
                        (default: 0)
  --molecular-dna-absolute MOLECULAR_DNA_ABSOLUTE
                        Strict DNA TPM threshold; requires dna-percent 0
  --molecular-rna-absolute MOLECULAR_RNA_ABSOLUTE
                        Strict RNA TPM threshold; requires rna-percent 0
  --molecular-metabolites
                        Also select through positive compounds and direct reaction/KO
                        links
  --background BACKGROUND
                        Common target-only host/contaminant FASTA for molecular runs
  --annotation-settings ANNOTATION_SETTINGS
                        Optional JSON for local eggNOG/ESM-C annotation after grouping