Crate uno_core 

uno-core: MS1-only protein identification.

Assigns peptide/protein identities from single-run feature files or KOTH consensus matrices. Long-format bundles retain original feature quality and per-run observations for shared-database batch search.

Re-exports

pub use calibrate::Calibration;

pub use database::DigestConfig;

pub use database::PeptideDb;

pub use database::TheoPeptide;

pub use im::im_features;

pub use im::ImModel;

pub use im::IM_N_FEATURES;

pub use input::Feature;

pub use input::FeatureMatrix;

pub use matching::match_features;

pub use matching::match_features_topk;

pub use matching::MatchConfig;

pub use matching::Pfm;

pub use protein::ProteinConfig;

pub use protein::ProteinId;

pub use rt::rt_features;

pub use rt::RtModel;

pub use rt::RT_N_FEATURES;

pub use rt_cnn::RtCnn;

pub use rt_mlp::RtMlp;

pub use score::score;

pub use score::ScoreConfig;

pub use score::ScoreResult;

Modules

calibrate

Monotone piecewise-linear calibration of abstract RT -> observed RT. Ported from sherlock’s rt_model::Calibration: quantile-bin by abstract RT, take per-bin medians, force the observed knots non-decreasing, interpolate.

cell_evidence

Optional extraction descriptors. These are learned covariates, not identity probabilities or additional peptide observations. Zero-intensity cells are unfilled; an unavailable q-value never counts as a confident extraction.

database

Phase 2: in-silico peptide database.

decoy

Alternative decoy-protein generators (experimental).

digest

In-silico digestion: FASTA parsing, enzymatic cleavage, and static/variable modification expansion.

group

Protein grouping for group-aware ID and quant.

im

Charge-aware IM prediction with the shared RT sequence CNN or legacy tree models.

input

Read single-run biosaur2/KOTH feature TSVs or aligned KOTH intensity matrices.

joint

Candidate-specific evidence from the original observations of an aligned row. Measurements are features of one discriminant, NOT independent probabilities.

long_input

KOTH long-format bundle ingestion. Quantities are attached once per cell; original observations remain separately available for per-run batch search.

mass

High-precision (f64) monoisotopic masses.

matching

Phase 3: match single-run features to theoretical peptides by precursor mass.

protein

Phase 6: protein-level aggregation + FDR.

quant

Multi-sample quantitation over a koth-aligned matrix: correct the peptide->protein feature assignment using the across-sample intensity pattern, then emit a protein-group × sample abundance matrix.

rt

Retention-time prediction: shared featurization + a gbdt model distilled from DeepLC.

rt_cache

On-disk cache of abstract RT predictions.

rt_cnn

Pure-Rust inference for the URTC retention-time CNN — a sequence-aware model (learned AA embedding -> residual dilated Conv1d stack -> masked mean+max pool -> MLP head), trained offline by crates/uno-rt-train/python/train_rt.py on experimental retention times from the Chronologer peptide library (see crates/uno-rt-train/python/chronologer_to_labels.py). The shipped model is models/rt_cnn_chrono_global_v3_ch32.urtc; unlike the rt.rs student it is NOT distilled from DeepLC.

rt_mlp

Pure-Rust MLP retention-time model — a faster, smaller, more accurate student than the distilled gbdt (see rt.rs). Trained offline (Keras) on the shared crate::rt::rt_features and serialized to a compact binary (models/rt_mlp.bin).

score

Phase 5: discriminant scoring, bootstrap RT calibration, and target-decoy FDR.

Functions

seed_mix

Mix a user-chosen seed into the fixed constants behind uno’s stochastic stages: the global-pool decoy shuffle, RANSAC pair sampling, CV fold assignment, and the tree booster’s row/column subsampling. Seed 0 is the shipped default and returns 0, so every constant is left untouched and historical output is reproduced byte for byte; any other seed perturbs all of those stages together (uno --seed N). Used to measure uno’s seed-to-seed spread the same way ms1searchpy’s run-to-run spread is measured.