Skip to content

Remove unused decoy-group margin features and computation - #153

Merged
jspaezp merged 2 commits into
mainfrom
fix/remove-decoy-group-rescore-features
Oct 2, 2026
Merged

jspaezp merged 2 commits into
mainfrom
fix/remove-decoy-group-rescore-features

Conversation

@jspaezp

@jspaezp jspaezp commented Sep 29, 2026 •

Copy link
Copy Markdown
Collaborator

Change

Remove the two decoy-group margin features from all rescorer lanes and from the result schema. Competition now sorts by group key only and scans each group for its highest score; it no longer sorts members by score or computes runner-up margins. Sequence deduplication still sorts by chemistry and score to keep the best duplicate. Bump the results format to 6 because two Parquet columns were removed.

Entrapment result

The feature-removal trial on one HeLa run used the same 1,753,338 competed candidates and identical raw scores:

Metric Before After
Original-target peptide/charge IDs at empirical paired FDP ≤1% 23,048 31,012
Paired FDP at reported q≤1% 4.21% 1.29%

The subsequent competition-sort cleanup has not been benchmarked on that run. The trial is exploratory evidence, not a calibration claim. Trial notebook.

Checks

  • task fmt
  • cargo test -p timsseek -p timsseek_cli --lib (202 passed, 1 ignored)
  • cargo test -p timsseek_cli --bin timsseek (110 passed)
  • cargo clippy -p timsseek -p timsseek_cli --all-targets -- -D warnings

@github-actions

github-actions Bot commented Sep 29, 2026 •

Copy link
Copy Markdown

Apex-finder bench

cargo run -p apex_sim --release --example bench -- 1000 2

Sensitivity + timing across canonical scenarios
=== summary (n=1000, tol=±2 cycles) ===
  scenario                      pass2%  pass1%   medErr    us/run
  clean                          100.0   100.0        0     27.35
  moderate_noise                  92.7    92.7        0     27.31
  high_noise+interference         59.3    59.3        1     26.80
  heavy_interference               7.4     7.4       88     26.71
  mismatched_library              59.8    59.8        1     26.77
  absent_top_fragment             16.0    16.0       65     26.78
  absent_precursor                59.3    59.3        1     26.77

=== broad apex-finding (n=500, tol=±2 cycles) ===
  scenario                      pass2%  pass1%   medErr    us/run
  broad_clean                    100.0   100.0        0    162.90
  broad_moderate_noise            50.6    50.6        1    158.65
  broad_high_noise+interf         27.6    27.6      308    158.75
  broad_hard_3x_density            0.8     0.8      487    158.78
  broad_mismatched_library        41.0    41.0      161    158.39
  broad_measured_density          49.6    49.6       13    188.51

=== narrow recovery (n=1000, tol=±2 cycles) ===
  scenario                      pass2%  pass1%   medErr    us/run
  narrow_clean                   100.0   100.0        0     18.58
  narrow_moderate_noise           81.3    81.3        0     18.26
  narrow_high_noise+interf        65.8    65.8        1     18.24
  narrow_hard_3x_density          10.9    10.9       34     18.20
  narrow_mismatched_library       62.8    62.8        1     18.21
  narrow_measured_density         83.6    83.6        0     20.96

=== narrow score discrimination (AUC, n_seed_pairs=1000) ===
  scenario                         AUC   med+signal    med-noise
  narrow_clean                   1.000     4.329e14      7.199e7
  narrow_moderate_noise          0.870      4.054e9      3.054e8
  narrow_high_noise+interf       0.830      2.761e9      3.602e8
  narrow_hard_3x_density         0.684      3.644e9      1.244e9
  narrow_mismatched_library      0.842      2.172e9      3.575e8
  narrow_measured_density        0.858      4.341e9      2.562e8

commit a7180af

@jspaezp jspaezp changed the title Exclude decoy-group margins from rescorer features Remove unused decoy-group margin features and computation Sep 29, 2026
@jspaezp
jspaezp merged commit 3c0dc44 into main Oct 2, 2026
8 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant