PySingleCellNet/pySingleCellNet/scn_train.py at e1d31f241ddc5047b9e9511fa973a27201bfb440 · CahanLab/PySingleCellNet · GitHub

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
import numpy as np
import pandas as pd
import scanpy as sc
import anndata as ad
from sklearn.ensemble import RandomForestClassifier
import warnings
from .utils import *
from .tsp_rf import *

def randomize(expDat, num=50):
    temp=expDat.to_numpy()
    temp=np.array([np.random.choice(x, len(x), replace=False) for x in temp])
    temp=temp.T
    temp=np.array([np.random.choice(x, len(x), replace=False) for x in temp]).T
    return pd.DataFrame(data=temp, columns=expDat.columns).iloc[0:num,:]

def sc_trans_rnaseq(aDat,total = 10000 ):
    sc.pp.normalize_per_cell(aDat, counts_per_cell_after=total)
    sc.pp.log1p(aDat)
    sc.pp.scale(aDat, max_value=10)
    #return np.log(1+expCountDnW)
    return aDat

def sc_makeClassifier(expTrain, genes, groups, nRand=70, ntrees=2000, stratify=False):
    randDat = randomize(expTrain, num=nRand)
    expT = pd.concat([expTrain, randDat])
    allgenes = expT.columns.values
    missingGenes = np.setdiff1d(np.unique(genes), allgenes)
    ggenes= np.intersect1d(np.unique(genes), allgenes)
    if not stratify:
        clf = RandomForestClassifier(n_estimators=ntrees, random_state=100)
    else:
        clf = RandomForestClassifier(n_estimators=ntrees,class_weight="balanced", random_state=100)
    ggroups=np.append(np.array(groups), np.repeat("rand", nRand)).flatten()
    clf.fit(expT.loc[:,ggenes].to_numpy(), ggroups)
    return clf

def scn_train(aTrain,dLevel,nTopGenes = 100,nTopGenePairs = 100,nRand = 100, nTrees = 1000,stratify=False,counts_per_cell_after=1e4, scaleMax=10, limitToHVG=False, normalization = True, include_all_genes = False):
    warnings.filterwarnings('ignore')
    stTrain= aTrain.obs

    expRaw = aTrain.to_df()
    expRaw = expRaw.loc[stTrain.index.values]

    adNorm = aTrain.copy()
    if normalization == True:
        sc.pp.normalize_per_cell(adNorm, counts_per_cell_after=counts_per_cell_after)
        sc.pp.log1p(adNorm)

        print("HVG")
        if limitToHVG:
            sc.pp.highly_variable_genes(adNorm, min_mean=0.0125, max_mean=4, min_disp=0.5)
            adNorm = adNorm[:, adNorm.var.highly_variable]

        sc.pp.scale(adNorm, max_value=scaleMax)

    expTnorm = adNorm.to_df()
    expTnorm=expTnorm.loc[stTrain.index.values]

    ### expTnorm= pd.DataFrame(data=aTrain.X,  index= aTrain.obs.index.values, columns= aTrain.var.index.values)
    ### expTnorm=expTnorm.loc[stTrain.index.values]
    print("Matrix normalized")
    ### cgenesA, grps, cgenes_list =findClassyGenes(expTnorm,stTrain, dLevel = dLevel, topX = nTopGenes)
    if include_all_genes == False:
        cgenesA, grps, cgenes_list =findClassyGenes(expTnorm,stTrain, dLevel = dLevel, topX = nTopGenes)
    else:
        cgenesA = np.array(aTrain.var.index)
        grps = aTrain.obs[dLevel]
        cgenes_list = dict()
        for g in np.unique(grps):
            cgenes_list[g] = cgenesA

    print("There are ", len(cgenesA), " classification genes\n")
    ### xpairs= ptGetTop(expTnorm.loc[:,cgenesA], grps, cgenes_list, topX=nTopGenePairs, sliceSize=5000)
    xpairs= ptGetTop(expTnorm.loc[:,cgenesA], grps, cgenes_list, topX=nTopGenePairs, sliceSize=5000)

    print("There are", len(xpairs), "top gene pairs\n")
    pdTrain= query_transform(expRaw.loc[:,cgenesA], xpairs)
    print("Finished pair transforming the data\n")
    tspRF=sc_makeClassifier(pdTrain.loc[:, xpairs], genes=xpairs, groups=grps, nRand = nRand, ntrees = nTrees, stratify=stratify)
    return [cgenesA, xpairs, tspRF, cgenes_list]

def scn_classify(adata, cgenes, xpairs, rf_tsp, nrand = 0 ):
    classRes = scn_predict(cgenes, xpairs, rf_tsp, adata, nrand = nrand)
    categories = classRes.columns.values
    adNew = ad.AnnData(classRes, obs=adata.obs, var=pd.DataFrame(index=categories))
    # adNew.obs['category'] =  classRes.idxmax(axis=1)
    adNew.obs['SCN_class'] =  classRes.idxmax(axis=1)
    return adNew

def add_classRes(adata: AnnData, adClassRes, copy=False) -> AnnData:
    cNames = adClassRes.var_names
    for cname in cNames:
        adata.obs[cname] = adClassRes[:,cname].X
    # adata.obs['category'] = adClassRes.obs['category']
    adata.obs['SCN_class'] = adClassRes.obs['SCN_class']
    return adata if copy else None

def check_adX(adata: AnnData) -> AnnData:
    from scipy import sparse
    if( isinstance(adata.X, np.ndarray)):
        adata.X = sparse.csr_matrix(adata.X)


def scn_predict(cgenes, xpairs, rf_tsp, aDat, nrand = 2):
    if isinstance(aDat.X,np.ndarray):
        # in the case of aDat.X is a numpy array
        aDat.X = ad._core.views.ArrayView(aDat.X)
###    expDat= pd.DataFrame(data=aDat.X, index= aDat.obs.index.values, columns= aDat.var.index.values)
    expDat= pd.DataFrame(data=aDat.X.toarray(), index= aDat.obs.index.values, columns= aDat.var.index.values)
    expValTrans=query_transform(expDat.reindex(labels=cgenes, axis='columns', fill_value=0), xpairs)
    classRes_val=rf_classPredict(rf_tsp, expValTrans, numRand=nrand)
    return classRes_val

def rf_classPredict(rfObj,expQuery,numRand=50):
    if numRand > 0 :
        randDat=randomize(expQuery, num=numRand)
        expQuery=pd.concat([expQuery, randDat])
    xpreds= pd.DataFrame(rfObj.predict_proba(expQuery), columns= rfObj.classes_, index=expQuery.index)
    return xpreds

def scn_train_edit(aTrain,dLevel,nTopGenes = 100,nTopGenePairs = 100,nRand = 100, nTrees = 1000,stratify=False,counts_per_cell_after=1e4, scaleMax=10, limitToHVG=True, normalization = True, include_all_genes = False):
    warnings.filterwarnings('ignore')
    stTrain= aTrain.obs

    expRaw = aTrain.to_df()
    expRaw = expRaw.loc[stTrain.index.values]

    adNorm = aTrain.copy()
    if normalization == True:
        sc.pp.normalize_per_cell(adNorm, counts_per_cell_after=counts_per_cell_after)
        sc.pp.log1p(adNorm)

        print("HVG")
        if limitToHVG:
            sc.pp.highly_variable_genes(adNorm, min_mean=0.0125, max_mean=4, min_disp=0.5)
            adNorm = adNorm[:, adNorm.var.highly_variable]

        sc.pp.scale(adNorm, max_value=scaleMax)

    expTnorm = adNorm.to_df()
    expTnorm=expTnorm.loc[stTrain.index.values]

    ### expTnorm= pd.DataFrame(data=aTrain.X,  index= aTrain.obs.index.values, columns= aTrain.var.index.values)
    ### expTnorm=expTnorm.loc[stTrain.index.values]
    print("Matrix normalized")
    ### cgenesA, grps, cgenes_list =findClassyGenes(expTnorm,stTrain, dLevel = dLevel, topX = nTopGenes)
    if include_all_genes == False:
        cgenesA, grps, cgenes_list =findClassyGenes_edit(adNorm, dLevel = dLevel, topX = nTopGenes)
    else:
        cgenesA = np.array(aTrain.var.index)
        grps = aTrain.obs[dLevel]
        cgenes_list = dict()
        for g in np.unique(grps):
            cgenes_list[g] = cgenesA

    print("There are ", len(cgenesA), " classification genes\n")
    ### xpairs= ptGetTop(expTnorm.loc[:,cgenesA], grps, cgenes_list, topX=nTopGenePairs, sliceSize=5000)
    xpairs= ptGetTop(expTnorm.loc[:,cgenesA], grps, cgenes_list, topX=nTopGenePairs, sliceSize=5000)

    print("There are", len(xpairs), "top gene pairs\n")
    pdTrain= query_transform(expRaw.loc[:,cgenesA], xpairs)
    print("Finished pair transforming the data\n")
    tspRF=sc_makeClassifier(pdTrain.loc[:, xpairs], genes=xpairs, groups=grps, nRand = nRand, ntrees = nTrees, stratify=stratify)
    return [cgenesA, xpairs, tspRF, cgenes_list]