BayesClassSpecificFDR.py

#!/usr/bin/env python3

import sys
import os
import getopt
import numpy as np
import numpy.polynomial.polynomial as poly
import matplotlib.pyplot as plt
'''
sort the file according to SpecEvalue in ascending order first
the script used the method described in the following paper to estimate subgroup specific FDR. 
Transferred subgroup false discovery rate for rare post-translational modifications detected by mass spectrometry.
Mol Cell Proteomics. 2014 May
'''

novel_prefix = "class_"
decoy_prefix = "XXX_class"
psm_qval = 0.05

if len(sys.argv[1:])<=1:  ### Indicates that there are insufficient number of command-line arguments
    print("Warning! wrong command!")
    print("Example: python BayesClassSpecificFDR.py --input PSM_filename --output output_filename --novel_prefix SAAV_ --decoy_prefix XXX_SAAV --psm_qval 0.05")
else:
    options, remainder = getopt.getopt(sys.argv[1:],'', ['input=','output=','novel_prefix=','decoy_prefix=','psm_qval='])
    for opt, arg in options:
        if opt == '--input': input_file=arg
        elif opt == '--output': output_file=arg 
        elif opt == '--decoy_prefix': decoy_prefix=arg
        elif opt == '--novel_prefix': novel_prefix=arg
        elif opt == '--psm_qval': psm_qval = float(arg)
        else:
            print("Warning! Command-line argument: %s not recognized. Exiting..." % opt); sys.exit()


input=open(input_file,'r')# the input tsv file need to be sorted by SpectEval in assending order first
output=open(output_file,'w')

header = input.readline().strip().split("\t")
header += ["ClassPSM-FDR","ClassPep-FDR"]

output.write("\t".join(header)+"\n")

score_dic = {}
decoy_dic={}

novel_targetcount=0
novel_decoycount=0

decoycount=0
targetcount=0

novpep_dic={}

pep_col=header.index("Peptide")
prot_col=header.index("Protein")

specEval_col=header.index("SpecEValue")

for line in input:
    row=line.strip().split('\t')
    pro=row[prot_col]
    specEval = -np.log10(float(row[specEval_col]))
    
    if "XXX_" in pro:
        decoycount+=1
        if decoy_prefix in pro:
            novel_decoycount+=1
        decoy_dic[specEval] = [novel_decoycount,decoycount]
    else:
        targetcount+=1
        if novel_prefix in pro:
            novel_targetcount+=1
 
    score_dic[specEval] = [targetcount,decoycount,novel_targetcount,novel_decoycount]


input.close()

x=[]
y=[]
x_filter = []
y_filter = []

for score in decoy_dic:
    x.append(score)
    frac=float(decoy_dic[score][0]/decoy_dic[score][1])
    y.append(frac)
    if 6<score<10:
        x_filter.append(score)
        y_filter.append(frac)
    
coefs = poly.polyfit(np.array(x_filter),np.array(y_filter), 1)
print("coefficients",coefs)
fit = poly.polyval(x, coefs)

fig=plt.figure()
ax = fig.add_subplot(111) 
ax.scatter(x,y,label = "Real data",s=1)
ax.plot(x,fit,label = "Polynomial with order=1", color='C1')
ax.legend()
plt.xlabel('-log10(SpecEValue)')
plt.ylabel('Gamma (class specific decoy hits / total decoy hits)')
fig.savefig("fitcurve.png")

input2=open(input_file,'r')

for line in input2:
    row=line.strip().split('\t')
    pep=row[pep_col]
    pro=row[prot_col]
    if novel_prefix not in pro:
        continue;
    
    specEval = -np.log10(float(row[specEval_col]))
    counts = score_dic[specEval]   

    targetcount=float(counts[0])
    decoycount=float(counts[1])
    FDR=decoycount/targetcount

    novel_targetcount=float(counts[2])
    gamma = poly.polyval(specEval, coefs)
    novelFDR = FDR*gamma*(targetcount/novel_targetcount)

    if pep not in novpep_dic:
        novpep_dic[pep]= novelFDR
   
    row.append(str(novelFDR))
    row.append(str(novpep_dic[pep]))
   
    if novelFDR < psm_qval:
        if "XXX_" not in pro: #write only target PSMs
            output.write("\t".join(row)+"\n")

print ("Hits in novel search space: targe,decoy",novel_targetcount,novel_decoycount)

input2.close()
output.close()