Fix pLDDT: fetch per-residue scores from AlphaFold confidence JSON endpoint (alphafold_client.py)

alphafold_client.py

@@ -129,14 +129,41 @@ def extract_protein_metadata(uniprot_data: dict) -> dict[str, Any]:
     }
 
 
-def extract_plddt_scores(alphafold_data: dict) -> list[float]:
+async def fetch_plddt_scores(alphafold_data: dict) -> list[float]:
     """
-    Extract per-residue pLDDT scores from AlphaFold summary.
-    Falls back to a single average value if per-residue data unavailable.
+    Fetch per-residue pLDDT scores from AlphaFold confidence JSON file.
+    The summary endpoint doesn't include scores inline — they're in a separate file.
+    Falls back gracefully if unavailable.
     """
-    scores = alphafold_data.get("confidenceScores") or []
-    if not scores:
-        avg = alphafold_data.get("confidenceAvgLocalScore") or \
-              alphafold_data.get("plddt") or 0.0
-        return [float(avg)]
-    return [float(s) for s in scores]
+    # Try the confidence JSON URL directly
+    conf_url = alphafold_data.get("confidenceUrl") or alphafold_data.get("confidence_url")
+    if not conf_url:
+        # Construct from entry ID and version
+        entry_id = alphafold_data.get("entryId", "")
+        version = alphafold_data.get("latestVersion", 4)
+        if entry_id:
+            conf_url = f"https://alphafold.ebi.ac.uk/files/{entry_id}-confidence_v{version}.json"
+
+    if conf_url:
+        try:
+            async with httpx.AsyncClient(timeout=TIMEOUT) as client:
+                r = await client.get(conf_url)
+                r.raise_for_status()
+                data = r.json()
+                scores = data.get("confidenceScore") or data if isinstance(data, list) else []
+                if scores:
+                    return [float(s) for s in scores]
+        except Exception:
+            pass
+
+    # Final fallback: single average value from summary
+    avg = (alphafold_data.get("confidenceAvgLocalScore") or
+           alphafold_data.get("plddt") or 75.0)
+    return [float(avg)]
+
+
+def extract_plddt_scores(alphafold_data: dict) -> list[float]:
+    """Synchronous fallback — use fetch_plddt_scores() for accurate per-residue data."""
+    avg = (alphafold_data.get("confidenceAvgLocalScore") or
+           alphafold_data.get("plddt") or 75.0)
+    return [float(avg)]