#!/usr/bin/env python3 """Frozen entry023; exact-sequence checks precede motif-spacing regression.""" import argparse,json,os,re from pathlib import Path import numpy as np,pandas as pd ROOT=Path(__file__).resolve().parents[1] INPUT=Path(os.getenv('BIO_DISCOVERY_CACHE','/mnt/data/research/bioinformatics-discovery/cache/round-003'))/'voichek2024_tables.xlsx' def sites(s):return {m.start() for m in re.finditer(r'(?=[CT][ACG]GATC[CGT][AG])',s)} def fit(d,y): x=np.column_stack([np.ones(len(d)),pd.get_dummies(d.parent.astype(str),drop_first=True,dtype=float),pd.get_dummies(d.step,drop_first=True,dtype=float),d[['delta_gc','position','crowded']]]) b=np.linalg.lstsq(x,y,rcond=None)[0];u=y-x@b;bread=np.linalg.pinv(x.T@x);g=pd.DataFrame(x*u[:,None]).groupby(d.parent.to_numpy()).sum().to_numpy();rank=np.linalg.matrix_rank(x);cov=bread@(g.T@g)@bread*len(g)/(len(g)-1)*(len(d)-1)/(len(d)-rank);se=float(np.sqrt(cov[-1,-1]));return {'effect':float(b[-1]),'cluster_se':se,'t':float(b[-1]/se),'rank':int(rank)} def main(): global INPUT,ROOT ap=argparse.ArgumentParser();ap.add_argument('--input',type=Path,default=INPUT);ap.add_argument('--output',type=Path,default=ROOT/'results');args=ap.parse_args();INPUT=args.input;out=args.output;out.mkdir(parents=True,exist_ok=True) native=pd.read_excel(INPUT,sheet_name='S3 - MPRA results mix 1',usecols=[0,1,2,3]);native=native[native.Origin.str.match(r'^AT1G\d{5}$',na=False)].set_index('Sequence index OP1') d=pd.read_excel(INPUT,sheet_name='S4 - MPRA results mix 2');d=d[d['Sequence index OP1'].isin(native.index)&d.modification.str.startswith('add_',na=False)].copy() d.Sequence=d.Sequence.str.upper();rows=[];exclusions={} for parent,group in d.groupby('Sequence index OP1'): bymod=group.set_index('modification');original=native.loc[parent,'Sequence'].upper() for _,r in group.iterrows(): parts=r.modification.split('_');step=len(parts)-1;prevmod='_'.join(parts[:-1]);previous=original if step==1 else (bymod.loc[prevmod,'Sequence'] if prevmod in bymod.index else None) if previous is None:continue old,new=sites(previous),sites(r.Sequence);added=new-old reason='ok' if len(previous)!=160 or len(r.Sequence)!=160:reason='length' elif old-new or len(added)!=1:reason='not_exactly_one_net_motif' elif not old:reason='no_previous_motif' exclusions[reason]=exclusions.get(reason,0)+1 if reason!='ok':continue pos=next(iter(added));distance=min(abs(pos-i) for i in old) row={'parent':int(parent),'gene':native.loc[parent,'Origin'],'step':step,'old_motifs':len(old),'new_position':pos,'nearest_distance':distance,'crowded':int(distance<=16),'delta_gc':sum(r.Sequence.count(b)-previous.count(b) for b in 'GC'),'position':pos/160} for host in ['35S','pTRP1']: a=r['modified_'+host];b=r['unmodified_'+host] if step==1 else bymod.loc[prevmod,'modified_'+host] row['delta_'+host]=pd.to_numeric(a,errors='coerce')-pd.to_numeric(b,errors='coerce') rows.append(row) f=pd.DataFrame(rows).dropna().reset_index(drop=True);f.to_csv(out/'023_pilot_steps.csv',index=False) s={'eligible_steps':len(f),'parents':f.parent.nunique(),'crowded':int(f.crowded.sum()),'uncrowded':int((1-f.crowded).sum()),'sequence_exclusions':exclusions,'advance':False} if s['parents']>=40 and s['eligible_steps']>=120 and min(s['crowded'],s['uncrowded'])>=25: s['models']={h:fit(f,f['delta_'+h].to_numpy()) for h in ['35S','pTRP1']};s['advance']=all(v['effect']<=-.15 and v['t']<=-3 for v in s['models'].values());s['decision']='Advance to novelty search.' if s['advance'] else 'Frozen two-backbone effect gate failed.' else:s['decision']='Frozen eligibility gate failed.' (out/'023_summary.json').write_text(json.dumps(s,indent=2)+'\n');print(json.dumps(s,indent=2)) if __name__=='__main__':main()