Software runs and manuscript reproduction#
A software example, a figure rebuild and a replay of the original scientific analysis answer different questions. Choose the level you need.
| Goal | Starting point | What to retain |
|---|---|---|
| Learn or test the supported workflow | CAMPI example | Example validation and command logs |
| Analyse a new study | Acquisition manifest and search settings | Exact inputs, settings, executable versions and output provenance |
| Reproduce a manuscript result | The manuscript version's source-data package and producing analysis code | Manuscript identity, original executable/configuration identities, input and output hashes |
The companion paper repository
is maintained separately and may require collaborator access. Its
manuscript/CURRENT.json identifies its packaged manuscript; verify that it
matches the document you are reading. A newer manuscript circulated separately
does not automatically update that repository. Historical plotting profiles
reproduce their declared snapshots.
Distinguish the peptide populations#
For the September manuscript workflow:
| Operation | Convention | Manifest-runner option |
|---|---|---|
| Exact lookup | Eligible predictions of 9–50 residues; top 30% by score, including cutoff ties; I/L equivalent | --min-length 9 --max-length 50 --top-fraction 0.30 |
| Selection parsimony | All eligible acquisition predictions mapped onto candidates; fixed evidence counts; accession-hash ties | Runner uses hash-acc |
| Main search | Search peptides of 5–50 residues; subsequent target peptide q ≤ 0.01 | --search-min-length 5 --search-max-length 50 |
| Seven-residue sensitivity search | A separate search with confidence re-estimated | --search-min-length 7, with the remaining archived settings |
| Nested 9–50 analysis | Subset of accepted peptides from the original 5–50 search, retaining that search's q values | A downstream subset, not a new minimum-9 search |
These are selected conventions, not a complete reproduction command. Use the exact archived reservoir, prediction exports, acquisition roster and search configuration for each comparison. Support for another predictor's input format does not establish equivalent scientific performance.
Preserve the producing software#
Ordinary identification searches and corrected joint-LFQ analyses have distinct Sage provenance. Use the matching source, patch and build records for the particular result; the bundled example executable does not stand in for every historical producing build. Experimental bag/substitution analyses also have their own producing code and are separate from the supported exact-lookup route.
In a supplied paper checkout, python paper.py verify checks packaged identities
and declared producer coverage. Follow that version's instructions for numerical
recomputation or figure generation, using a new output directory. Byte integrity,
saved-result arithmetic and full raw-data replay are distinct checks.
The validation overview explains how these checks relate to scientific interpretation.