File size: 7,262 Bytes
3290550 |
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 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 |
import sys
import click
import numpy as np
from sklearn.neighbors import KDTree
import pandas as pd
from tqdm import tqdm
def rhoDelta(data,resol,dc,radius):
pos = data[[1, 4]].to_numpy() // resol
posTree = KDTree(pos, leaf_size=30, metric='chebyshev')
NNindexes, NNdists = posTree.query_radius(pos, r=radius, return_distance=True)
_l = []
for v in NNindexes:
_l.append(len(v))
_l=np.asarray(_l)
data = data[_l>5].reset_index(drop=True)
if data.shape[0] != 0:
pos = data[[1, 4]].to_numpy() // resol
val = data[6].to_numpy()
try:
posTree = KDTree(pos, leaf_size=30, metric='chebyshev')
NNindexes, NNdists = posTree.query_radius(pos, r=dc, return_distance=True)
except ValueError as e:
if "Found array with 0 sample(s)" in str(e):
print("#"*88,'\n#')
print("#\033[91m Error!!! The data is too sparse. Please decrease the value of: [t]\033[0m\n#")
print("#"*88,'\n')
sys.exit(1)
else:
raise
rhos = []
for i in range(len(NNindexes)):
rhos.append(np.dot(np.exp(-(NNdists[i] / dc) ** 2), val[NNindexes[i]]))
rhos = np.asarray(rhos)
_r = 100
_indexes, _dists = posTree.query_radius(pos, r=_r, return_distance=True, sort_results=True)
deltas = rhos * 0
LargerNei = rhos * 0 - 1
for i in range(len(_indexes)):
idx = np.argwhere(rhos[_indexes[i]] > rhos[_indexes[i][0]])
if idx.shape[0] == 0:
deltas[i] = _dists[i][-1] + 1
else:
LargerNei[i] = _indexes[i][idx[0]]
deltas[i] = _dists[i][idx[0]]
failed = np.argwhere(LargerNei == -1).flatten()
while len(failed) > 1 and _r < 100000:
_r = _r * 10
_indexes, _dists = posTree.query_radius(pos[failed], r=_r, return_distance=True, sort_results=True)
for i in range(len(_indexes)):
idx = np.argwhere(rhos[_indexes[i]] > rhos[_indexes[i][0]])
if idx.shape[0] == 0:
deltas[failed[i]] = _dists[i][-1] + 1
else:
LargerNei[failed[i]] = _indexes[i][idx[0]]
deltas[failed[i]] = _dists[i][idx[0]]
failed = np.argwhere(LargerNei == -1).flatten()
data['rhos']=rhos
data['deltas']=deltas
else:
data['rhos']=[]
data['deltas']=[]
return data
@click.command()
@click.option('-dc','--distance_cutoff', type=int, default=5, help='Distance cutoff for local density calculation in terms of bin. [5]')
@click.option('-t','--threshold', type=float, default=0.6, help='Loop score threshold [0.6]')
@click.option('-r','--resol', default=5000, help='resolution [5000]')
@click.option('-R','--radius', type=int, default=2, help='Radius threshold to remove outliers. [2]')
@click.option('-d','--mindelta', type=float, default=5, help='Min distance allowed between two loops [5]')
@click.option('-i','--candidates', type=str,required=True,help ='Loop candidates file path')
@click.option('-o','--output', type=str,required=True,help ='.bedpe file path to save loops')
def pool(distance_cutoff,candidates,resol,mindelta,threshold,output,radius,refine=True):
"""Call loops from loop candidates by clustering
"""
print('\npolaris loop pool START :) ')
data = pd.read_csv(candidates, sep='\t', header=None)
ccs = set(data.iloc[:,0])
if data.shape[0] == 0:
print("#"*88,'\n#')
print("#\033[91m Error!!! The file is empty. Please check your file.\033[0m\n#")
print("#"*88,'\n')
sys.exit(1)
data = data[data[6] > threshold].reset_index(drop=True)
data = data[data[4] - data[1] > 11*resol].reset_index(drop=True)
if data.shape[0] == 0:
print("#"*88,'\n#')
print("#\033[91m Error!!! The data is too sparse. Please decrease: [threshold] (minimum: 0.5).\033[0m\n#")
print("#"*88,'\n')
sys.exit(1)
data[['rhos','deltas']]=0
data=data.groupby([0]).apply(rhoDelta,resol=resol,dc=distance_cutoff,radius=radius).reset_index(drop=True)
minrho=0
targetData=data.reset_index(drop=True)
loopPds=[]
chroms=tqdm(set(targetData[0]), dynamic_ncols=True)
for chrom in chroms:
chroms.desc = f"[Runing clustering on {chrom}]"
data = targetData[targetData[0]==chrom].reset_index(drop=True)
pos = data[[1, 4]].to_numpy() // resol
posTree = KDTree(pos, leaf_size=30, metric='chebyshev')
rhos = data['rhos'].to_numpy()
deltas = data['deltas'].to_numpy()
centroid = np.argwhere((rhos > minrho) & (deltas > mindelta)).flatten()
_r = 100
_indexes, _dists = posTree.query_radius(pos, r=_r, return_distance=True, sort_results=True)
LargerNei = rhos * 0 - 1
for i in range(len(_indexes)):
idx = np.argwhere(rhos[_indexes[i]] > rhos[_indexes[i][0]])
if idx.shape[0] == 0:
pass
else:
LargerNei[i] = _indexes[i][idx[0]]
failed = np.argwhere(LargerNei == -1).flatten()
while len(failed) > 1 and _r < 100000:
_r = _r * 10
_indexes, _dists = posTree.query_radius(pos[failed], r=_r, return_distance=True, sort_results=True)
for i in range(len(_indexes)):
idx = np.argwhere(rhos[_indexes[i]] > rhos[_indexes[i][0]])
if idx.shape[0] == 0:
pass
else:
LargerNei[failed[i]] = _indexes[i][idx[0]]
failed = np.argwhere(LargerNei == -1).flatten()
LargerNei = LargerNei.astype(int)
label = LargerNei * 0 - 1
for i in range(len(centroid)):
label[centroid[i]] = i
decreasingsortedIdxRhos = np.argsort(-rhos)
for i in decreasingsortedIdxRhos:
if label[i] == -1:
label[i] = label[LargerNei[i]]
val = data[6].to_numpy()
refinedLoop = []
label = label.flatten()
for l in set(label):
idx = np.argwhere(label == l).flatten()
if len(idx) > 0:
refinedLoop.append(idx[np.argmax(val[idx])])
if refine:
loopPds.append(data.loc[refinedLoop])
else:
loopPds.append(data.loc[centroid])
loopPd=pd.concat(loopPds).sort_values(6,ascending=False)
loopPd[[1, 2, 4, 5]] = loopPd[[1, 2, 4, 5]].astype(int)
loopPd[[0,1,2,3,4,5,6]].to_csv(output,sep='\t',header=False, index=False)
ccs_ = set(loopPd.iloc[:,0])
badc = ccs.difference(ccs_)
if len(badc) == len(ccs):
raise ValueError("polaris loop pool FAILED :(\nPlease check input and mcool file to yield scoreFile. Or use higher '-s' value for more sparse mcool data.")
else:
print(f'\npolaris loop pool FINISHED :)\n{len(loopPd)} loops saved to {output}')
if len(badc) > 0:
print(f"But the loop score of {badc} are too sparse.\nYou may need to check the mcool data or re-run polaris loop score by increasing -s.")
if __name__ == '__main__':
pool()
|