| """Label-preserving dataset normalization and connected-component splitting.""" |
| from collections import defaultdict |
| import hashlib,json |
|
|
| LABELS={'positive':1,'negative':0,'unknown':None,'pos':1,'neg':0,'1':1,'0':0} |
|
|
| def normalize_snooppi(row): |
| raw=str(row['SNOOPPI_final_label']).lower().strip() |
| if raw not in LABELS:raise ValueError('unrecognized SNOOPPI label: '+raw) |
| return {'label':LABELS[raw],'label_source':raw,'partner_a':row['partner_A_sequence'], |
| 'partner_b':row['partner_B_sequence'],'source':row.get('pubmed_ids',''), |
| 'publication_years':row.get('publication_years',''),'assays':row.get('assays','')} |
|
|
| def connected_splits(rows,seed=2027): |
| """Union target clusters, peptide clusters, and publication IDs across rows. |
| |
| Cluster IDs are computed before this function (e.g. MMseqs2). All rows |
| sharing any entity or primary publication receive the same split. |
| """ |
| parent=list(range(len(rows))) |
| def root(i): |
| while parent[i]!=i:parent[i]=parent[parent[i]];i=parent[i] |
| return i |
| def union(i,j): |
| a,b=root(i),root(j) |
| if a!=b:parent[max(a,b)]=min(a,b) |
| seen={} |
| for i,r in enumerate(rows): |
| groups=['target:'+v for v in r['target_clusters']]+['peptide:'+v for v in r['peptide_clusters']]+['source:'+v for v in r['source_ids']] |
| if not groups:raise ValueError('missing split source records') |
| for g in groups: |
| if g in seen:union(i,seen[g]) |
| else:seen[g]=i |
| members=defaultdict(list) |
| for i,r in enumerate(rows):members[root(i)].append(str(r['id'])) |
| labels={} |
| for k,ids in members.items(): |
| h=int(hashlib.sha256((str(seed)+'|'+ '|'.join(sorted(ids))).encode()).hexdigest()[:12],16)/16**12 |
| labels[k]='train' if h<.7 else 'ensemble_fit' if h<.8 else 'calibration' if h<.9 else 'test' |
| return [labels[root(i)] for i in range(len(rows))] |
|
|
