File size: 3,948 Bytes
1fa45cf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
import sys
from pathlib import Path
import h5py
import numpy as np
import pandas as pd

out = Path(sys.argv[2])
if sys.argv[1] == 'remote':
    from remote_h5 import Remote
    source = Remote('https://datasets.cellxgene.cziscience.com/a9d92e38-9a6e-401b-8484-74bb15122341.h5ad',
                    1463737857, sys.argv[3])
else:
    source = sys.argv[1]

def decode(node):
    if isinstance(node, h5py.Group):
        code = node['codes'][:]
        return np.append(node['categories'].asstr()[:], '')[code]
    return node.asstr()[:] if node.dtype.kind in 'OS' else node[:]

with h5py.File(source) as f:
    obs = f['obs']
    columns = ['HTAN_Participant_ID','HTAN_Biospecimen_ID','batch','histo','tissue','treatment',
               'cell_type_med','cell_type_general','procedure']
    cells = pd.DataFrame({key: decode(obs[key]) for key in columns})
    cells.columns = ['patient','sample','batch','histo','tissue','treatment','label','general','procedure']
    cells['group'] = np.where(cells.label.str.startswith('SCLC'), 'malignant',
                             np.where(cells.general.eq('Immune'),'immune',
                                      np.where(cells.label.eq('Epithelial Stroma'),'nonmalignant_epithelial',
                                               np.where(cells.label.isin(['Endothelial','Fibroblast']),cells.label,'other'))))
    cells.groupby(list(cells.columns)).size().rename('cells').reset_index().to_csv(out/'chan_sample_inventory.csv',index=False)
    keep = cells.histo.eq('SCLC') & cells.tissue.eq('lung')
    selected = cells.loc[keep].copy()
    print(selected.groupby(['patient','sample','batch','treatment','procedure','group']).size().to_string(),flush=True)
    if '--strict-only' in sys.argv:
        selected = selected[selected.patient.isin(['HTA8_2003','HTA8_2010','HTA8_2013','HTA8_2018'])]
    if '--patient' in sys.argv:
        selected = selected[selected.patient.eq(sys.argv[sys.argv.index('--patient')+1])]
    if '--inventory' in sys.argv:
        sys.exit()
    var = f['raw/var']
    names = decode(var['feature_name']) if 'feature_name' in var else decode(f['var/feature_name'])
    target = np.flatnonzero(np.isin(names,['CIP2A','KIAA1524']))
    cycle = np.flatnonzero(np.isin(names,['MKI67','TOP2A','UBE2C','CDK1','CENPF','BIRC5']))
    assert len(target)==1 and len(cycle)>=4
    raw = f['raw/X']
    ptr = raw['indptr'][:]
    values = []
    for left in range(0,len(cells),256):
        right = min(left+256,len(cells))
        rows = selected.index[(selected.index>=left)&(selected.index<right)]
        if not len(rows):
            continue
        low, high = ptr[rows[0]], ptr[rows[-1]+1]
        idx, dat = raw['indices'][low:high],raw['data'][low:high]
        assert np.all(dat>=0) and np.all(dat==np.floor(dat))
        for row in rows:
            a,b = ptr[row]-low,ptr[row+1]-low
            ii,dd = idx[a:b],dat[a:b]
            values.append((row,float(dd.sum()),float(dd[ii==target[0]].sum()),int(np.isin(ii,cycle).sum())))
    selected = selected.join(pd.DataFrame(values,columns=['row','library','CIP2A','cycle_markers']).set_index('row'))
    selected['detected']=selected.CIP2A.gt(0)
    selected['cycling_proxy']=np.where(selected.cycle_markers.ge(2),'2+ markers','0–1 markers')
    strict = ['HTA8_2003','HTA8_2010','HTA8_2013','HTA8_2018']
    for tag,subset,keys in [('all',selected,['patient','group']),('cycle',selected,['patient','group','cycling_proxy']),('confirmed_primary',selected[selected.patient.isin(strict)],['patient','group'])]:
        result=subset.groupby(keys).agg(cells=('library','size'),umi=('library','sum'),CIP2A=('CIP2A','sum'),detected=('detected','sum')).reset_index()
        result['cpm']=result.CIP2A/result.umi*1e6
        result['detection']=result.detected/result.cells
        result['target_share']=result.CIP2A/result.groupby('patient').CIP2A.transform('sum')
        result.to_csv(out/f'chan_{tag}.csv',index=False)