  gate 1: PASS  schema, row counts, identifiers and numeric values all conform; 22 reported genes lie outside the reference analysis set
  gate 2: PASS  40 of 40 footprint libraries assigned correctly (need 38)
  gate 3: PASS  P-site offsets match the simulator on every shared read length
  gate 4: PASS  Spearman against the reference table is 0.962 (need 0.85)
  gate 5: FAIL  length-bias slope is +0.0537 per log10 coding-sequence length (allowed +/-0.05)
  gate 6: FAIL  median slope recomputed from gene_slopes.tsv over the reference analysis set is 0.825 (allowed 0.88 to 1.15; the true value is 0.99. Ignoring measurement error reports 0.63, ignoring library contamination 0.83). Over the submission's own gene set it would be 0.808
  gate 7: FAIL  35 of 40 footprint libraries have the right set of contaminating lines and 40 the right foreign share (at most 3 may be wrong): RPF_L008 sources ['CL01', 'CL11'] vs ['CL01']; RPF_L009 sources ['CL03', 'CL07'] vs ['CL07']; RPF_L028 sources ['CL02'] vs none; RPF_L039 sources ['CL02'] vs none; RPF_L040 sources ['CL09'] vs none

  F1 0.426  precision 0.390  recall 0.470 (39 of 83 true buffered genes found)
  rank correlation against the true slope: 0.172

  FULL PASS: False   partial credit 0.3501
