fix(predict): decouple organism inference from GO gating; trust dropdown; kill oxygen-binding FP

feat(identify): UniProt-grounded organism + auto-filled accession; calibrate organism confidence

Fix saliency all-zeros regression: detach ESM embeddings before grad computation

feat: FABLE rename + organism selector + saliency 3D viewer

fix: _ESM_DIM global scope fix for model info endpoint

deploy: Add GO term explanation panel, CAFA-6 predict script, and OBO def parsing

deploy: bug fix, unified metrics, home.js + home.css

v4: creator credit, stage canvases, no-cache fix

feat: explanation fields in /api/go_terms response

feat: add /api/go_terms lightweight endpoint for pipeline integration

feat: unified_35M_v1 — 35M ESM + GOA-enriched training, auto ESM-dim detection

deploy: feat: taxon auto-detection, UniProt annotation comparison, Platt+temp loading

deploy: fix: tighten mammal false positive rate (threshold floor + dynamic cap)

deploy: feat: feature-level importance viz with 3D structure coloring

v4.2: precision-tuned mammal thresholds (T=3.85), fix threshold mismatch

ProtFunc v4.1: gradient saliency on ESM-2, 3Dmol.js AF viewer, UniProt accession input

deploy: unified_v1 v4.0 ins=0.9555 mam=0.2908

fix: 8124-class MLB + out_dim detection — mammal_enriched.pth now loads correctly, insect recall 95%

feat: taxon routing + organism selector (insect/mammal/auto)

fix: threshold dict format, temperature calibration at inference

perf: remove 1.1 GB CSV from HF download list

fix: remove dead code + dead var in propagate_and_filter

fix+enhance: device global, GO depth/names, replaced_by remapping

deploy mammal_enriched v3.2: insect=0.9542 mammal=0.5715 CAFA5-evaluated

perf: ESM embedding cache + 1500aa limit, add research scripts

Fix compute_dynamic_cap: lower rel_gap threshold and abs_cap

fix: skip corrupted checkpoints instead of degrading silently

fix: resolve embedding dimensions issue for supp_res2 predictions

Update server.py: v3-fixed model priority, generalization API + UI

fix: revert to improved model, add top-30 prediction cap

feat: non-AA validation, DNA detection, /api/structure endpoint

feat: v3 model support, GO hierarchy, supplemented features, updated UI

Update server.py

Update server.py

fix: add client+server input validation and GO hierarchy filtering

fix: add client+server input validation and GO hierarchy filtering

feat: input validation,

fix mlb load order again

fix mlb load order again

fix mlb load order again

validate sequences, reject invalid AA with clear error message

validate sequences, reject invalid AA with clear error message

fix mlb load order

fix mlb load order, insect demos, updated GO names

canonical GO names, clickable GO links, demo sequences

fix all model loading issues

fix model arch to match baseline_res keys

pair RecoveredBaselineModel with baseline_res.pth

switch to baseline ESM-only model

revert to correct model architecture