LSHTM_analysis/scripts/data_extraction.py

1205 lines
47 KiB
Python
Executable file

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
'''
Created on Tue Aug 6 12:56:03 2019
@author: tanu
'''
# FIXME: include error checking to enure you only
# concentrate on positions that have structural info?
# FIXME: import dirs.py to get the basic dir paths available
#=======================================================================
# TASK: extract ALL <gene> matched mutations from GWAS data
# Input data file has the following format: each row = unique sample id
# id,country,lineage,sublineage,drtype,drug,dr_muts_col,other_muts_col...
# 0,sampleID,USA,lineage2,lineage2.2.1,Drug-resistant,0.0,WT,gene_match<wt>POS<mut>; pncA_c.<wt>POS<mut>...
# where multiple mutations and multiple mutation types are separated by ';'.
# We are interested in the protein coding region i.e mutation with the<gene>_'p.' format.
# This script splits the mutations on the ';' and extracts protein coding muts only
# where each row is a separate mutation
# sample ids AND mutations are NOT unique, but the COMBINATION (sample id + mutation) = unique
# NOTE
#drtype is renamed to 'resistance' in the 35k dataset
# output files: all lower case
# 0) <gene>_common_ids.csv
# 1) <gene>_ambiguous_muts.csv
# 2) <gene>_mcsm_snps.csv
# 3) <gene>_metadata.csv
# 4) <gene>_all_muts_msa.csv
# 5) <gene>_mutational_positons.csv
# FIXME
## Make all cols lowercase
## change WildPos: wild_pos
## Add an extra col: wild_chain_pos
## output df: <gene>_linking_df.csv
#containing the following cols
#1. Mutationinformation
#2. wild_type
#3. position
#4. mutant_type
#5. chain
#6. wild_pos
#7. wild_chain_pos
#=======================================================================
#%% load libraries
import os, sys
import re
import pandas as pd
#import numpy as np
import argparse
#=======================================================================
#%% homdir and curr dir and local imports
homedir = os.path.expanduser('~')
# set working dir
os.getcwd()
os.chdir(homedir + '/git/LSHTM_analysis/scripts')
os.getcwd()
# import aa dict
from reference_dict import my_aa_dict # CHECK DIR STRUC THERE!
from tidy_split import tidy_split
#=======================================================================
#%% command line args
arg_parser = argparse.ArgumentParser()
arg_parser.add_argument('-d', '--drug', help='drug name', default = None)
arg_parser.add_argument('-g', '--gene', help='gene name (case sensitive)', default = None) # case sensitive
args = arg_parser.parse_args()
#=======================================================================
#%% variable assignment: input and output paths & filenames
drug = args.drug
gene = args.gene
#drug = 'pyrazinamide'
#gene = 'pncA'
gene_match = gene + '_p.'
print('mut pattern for gene', gene, ':', gene_match)
nssnp_match = gene_match+'[A-Z]{3}[0-9]+[A-Z]{3}'
print('nsSNP for gene', gene, ':', nssnp_match)
wt_regex = gene_match.lower()+'(\w{3})'
print('wt regex:', wt_regex)
mut_regex = r'\d+(\w{3})$'
print('mt regex:', mut_regex)
pos_regex = r'(\d+)'
print('position regex:', pos_regex)
# building cols to extract
dr_muts_col = 'dr_mutations_' + drug
other_muts_col = 'other_mutations_' + drug
print('Extracting columns based on variables:\n'
, drug
, '\n'
, dr_muts_col
, '\n'
, other_muts_col
, '\n===============================================================')
#=======================================================================
#%% input and output dirs and files
#=======
# dirs
#=======
datadir = homedir + '/' + 'git/Data'
indir = datadir + '/' + drug + '/' + 'input'
outdir = datadir + '/' + drug + '/' + 'output'
#=======
# input
#=======
#in_filename_master_master = 'original_tanushree_data_v2.csv' #19k
in_filename_master = 'mtb_gwas_meta_v3.csv' #33k
infile_master = datadir + '/' + in_filename_master
print('Input file: ', infile_master
, '\n============================================================')
#=======
# output
#=======
# several output files: in respective sections at the time of outputting files
print('Output filename: in the respective sections'
, '\nOutput path: ', outdir
, '\n=============================================================')
#%%end of variable assignment for input and output files
#=======================================================================
#%% Read input file
master_data = pd.read_csv(infile_master, sep = ',')
# column names
#list(master_data.columns)
# extract elevant columns to extract from meta data related to the drug
if in_filename_master == 'original_tanushree_data_v2.csv':
meta_data = master_data[['id'
, 'country'
, 'lineage'
, 'sublineage'
, 'drtype' #19k only
, drug
, dr_muts_col
, other_muts_col]]
if in_filename_master == 'mtb_gwas_meta_v3.csv':
core_cols = ['id'
, 'country'
, 'country2'
, 'geographic_source'
, 'region'
, 'date'
, 'strain'
, 'lineage'
, 'sublineage' #drtype renamed to resistance
, 'resistance'
, 'location'
, 'host_body_site'
, 'environment_material'
, 'host_status'
, 'hiv_status'
, 'HIV_status'
, 'isolation_source']
variable_based_cols = [drug
, dr_muts_col
, other_muts_col]
cols_to_extract = core_cols + variable_based_cols
print('Extracting', len(cols_to_extract), 'columns from master data')
meta_data = master_data[cols_to_extract]
del(master_data, variable_based_cols, cols_to_extract)
print('Extracted meta data from filename:', in_filename_master
, '\nDim:', meta_data.shape)
# checks and results
total_samples = meta_data['id'].nunique()
print('RESULT: Total samples:', total_samples
, '\n===========================================================')
# counts NA per column
meta_data.isna().sum()
print('No. of NAs/column:' + '\n', meta_data.isna().sum()
, '\n===========================================================')
#
# glance
#meta_data.head()
#total_samples - NA pyrazinamide = ?
# 19K: 19265-6754 = 12511
# 33K: 33681 - 23823 = 9858
# equivalent of table in R
# drug counts: complete samples for OR calcs
meta_data[drug].value_counts()
print('RESULT: Sus and Res samples:\n', meta_data[drug].value_counts()
, '\n===========================================================')
#%%
# IMPORTANT sanity check:
# This is to find out how many samples have 1 and more than 1 mutation,so you
# can use it to check if your data extraction process for dr_muts
# and other_muts has worked correctly AND also to check the dim of the
# final formatted data.
# This will have: unique COMBINATION of sample id and <gene_match> mutations
#========
# First: counting <gene_match> mutations in dr_muts_col column
#========
print('Now counting WT &', nssnp_match, 'muts within the column:', dr_muts_col)
# drop na and extract a clean df
clean_df = meta_data.dropna(subset=[dr_muts_col])
# sanity check: count na
na_count = meta_data[dr_muts_col].isna().sum()
if len(clean_df) == (total_samples - na_count):
print('PASS: clean_df extracted: length is', len(clean_df)
, '\nNo.of NAs in', dr_muts_col, '=', na_count, '/', total_samples
, '\n==========================================================')
else:
sys.exit('FAIL: Could not drop NAs')
dr_gene_count = 0
wt = 0
id_dr = []
id2_dr = []
#nssnp_match_regex = re.compile(nssnp_match)
for i, id in enumerate(clean_df.id):
#print (i, id)
#id_dr.append(id)
#count_gene_dr = clean_df[dr_muts_col].iloc[i].count(gene_match) # can include stop muts
count_gene_dr = len(re.findall(nssnp_match, clean_df[dr_muts_col].iloc[i], re.IGNORECASE))
#print(count_gene_dr)
if count_gene_dr > 0:
id_dr.append(id)
if count_gene_dr > 1:
id2_dr.append(id)
#print(id, count_gene_dr)
dr_gene_count = dr_gene_count + count_gene_dr
count_wt = clean_df[dr_muts_col].iloc[i].count('WT')
wt = wt + count_wt
print('RESULTS:')
print('Total WT in dr_muts_col:', wt)
print('Total matches of', gene, 'SNP matches in', dr_muts_col, ':', dr_gene_count)
print('Total samples with > 1', gene, 'muts in dr_muts_col:', len(id2_dr) )
print('=================================================================')
del(clean_df, na_count, i, id, wt, id2_dr, count_gene_dr, count_wt)
#========
# Second: counting <gene_match> mutations in dr_muts_col column
#========
print('Now counting WT &', nssnp_match, 'muts within the column:', other_muts_col)
# drop na and extract a clean df
clean_df = meta_data.dropna(subset=[other_muts_col])
# sanity check: count na
na_count = meta_data[other_muts_col].isna().sum()
if len(clean_df) == (total_samples - na_count):
print('PASS: clean_df extracted: length is', len(clean_df)
, '\nNo.of NAs =', na_count, '/', total_samples
, '\n=========================================================')
else:
sys.exit('FAIL: Could not drop NAs')
other_gene_count = 0
wt_other = 0
id_other = []
id2_other = []
for i, id in enumerate(clean_df.id):
#print (i, id)
#id_other.append(id)
#count_gene_other = clean_df[other_muts_col].iloc[i].count(gene_match) # can include stop muts
count_gene_other = len(re.findall(nssnp_match, clean_df[other_muts_col].iloc[i], re.IGNORECASE))
if count_gene_other > 0:
id_other.append(id)
if count_gene_other > 1:
id2_other.append(id)
#print(id, count_gene_other)
other_gene_count = other_gene_count + count_gene_other
count_wt = clean_df[other_muts_col].iloc[i].count('WT')
wt_other = wt_other + count_wt
print('RESULTS:')
print('Total WT in other_muts_col:', wt_other)
print('Total matches of', gene, 'SNP matches in', other_muts_col, ':', other_gene_count)
print('Total samples with > 1', gene, 'muts in other_muts_col:', len(id2_other) )
print('=================================================================')
print('Predicting total no. of rows in the curated df:', dr_gene_count + other_gene_count
, '\n===================================================================')
expected_rows = dr_gene_count + other_gene_count
del(i, id, wt_other, clean_df, na_count, id2_other, count_gene_other, count_wt)
#%%
############
# extracting dr and other muts separately along with the common cols
#############
print('Extracting dr_muts from col:', dr_muts_col, 'with other meta_data')
print('muts to extract:', nssnp_match )
#===============
# dr mutations: extract gene_match entries with meta data and ONLY dr_muts col
#===============
if in_filename_master == 'original_tanushree_data_v2.csv':
meta_data_dr = meta_data[['id'
,'country'
,'lineage'
,'sublineage'
,'drtype'
, drug
, dr_muts_col]]
if in_filename_master == 'mtb_gwas_meta_v3.csv':
dr_based_cols = [drug, dr_muts_col]
cols_to_extract = core_cols + dr_based_cols
print('Extracting', len(cols_to_extract), 'columns from meta data')
meta_data_dr = meta_data[cols_to_extract]
del(dr_based_cols, cols_to_extract)
if meta_data_dr.shape[0] == len(meta_data) and meta_data_dr.shape[1] == (len(meta_data.columns)-1):
print('PASS: Dimensions match'
, '\n===============================================================')
else:
print('FAIL: Dimensions mismatch:'
, 'Expected dim:', len(meta_data), (len(meta_data.columns)-1)
, '\nGot:', meta_data_dr.shape
, '\n===============================================================')
sys.exit()
# Extract within this the nsSNPs for gene of interest using string match
#meta_gene_dr = meta_data_dr.loc[meta_data_dr[dr_muts_col].str.contains(gene_match, na = False)]
meta_gene_dr = meta_data_dr.loc[meta_data_dr[dr_muts_col].str.contains(nssnp_match, na = False, regex = True, case = False)]
print('gene_snp_match in dr:', len(meta_gene_dr))
dr_id = meta_gene_dr['id'].unique()
print('RESULT: No. of samples with dr muts in pncA:', len(dr_id))
if len(id_dr) == len(meta_gene_dr):
print('PASS: lengths match'
, '\n===============================================================')
else:
print('FAIL: length mismatch'
, '\nExpected len:', len(id_dr)
, '\nGot:', len(meta_gene_dr))
sys.exit()
dr_id = pd.Series(dr_id)
#=================
# other mutations: extract nssnp_match entries from other_muts_col
#==================
print('Extracting other_muts from:', other_muts_col,'with other meta_data')
print('muts to extract:', nssnp_match)
if in_filename_master == 'original_tanushree_data_v2.csv':
meta_data_other = meta_data[['id'
, 'country'
, 'lineage'
, 'sublineage'
, 'drtype'
, drug
, other_muts_col]]
if in_filename_master == 'mtb_gwas_meta_v3.csv':
other_based_cols = [drug, other_muts_col]
cols_to_extract = core_cols + other_based_cols
print('Extracting', len(cols_to_extract), 'columns from meta data')
meta_data_other = meta_data[cols_to_extract]
del(other_based_cols, cols_to_extract)
if meta_data_other.shape[0] == len(meta_data) and meta_data_other.shape[1] == (len(meta_data.columns)-1):
print('PASS: Dimensions match'
, '\n===============================================================')
else:
print('FAIL: Dimensions mismatch:'
, 'Expected dim:', len(meta_data), (len(meta_data.columns)-1)
, '\nGot:', meta_data_other.shape
, '\n===============================================================')
sys.exit()
# Extract within this nSSNP for gene of interest using string match
#meta_gene_other = meta_data_other.loc[meta_data_other[other_muts_col].str.contains(gene_match, na = False)]
meta_gene_other = meta_data_other.loc[meta_data_other[other_muts_col].str.contains(nssnp_match, na = False, regex = True, case = False)]
print('gene_snp_match in other:', len(meta_gene_other))
other_id = meta_gene_other['id'].unique()
print('RESULT: No. of samples with other muts:', len(other_id))
if len(id_other) == len(meta_gene_other):
print('PASS: lengths match'
, '\n==============================================================')
else:
print('FAIL: length mismatch'
, '\nExpected len:', len(id_other)
, '\nGot:', len(meta_gene_other))
sys.exit()
other_id = pd.Series(other_id)
#%% Find common IDs
print('Now extracting common_ids...')
common_mut_ids = dr_id.isin(other_id).sum()
print('RESULT: No. of common ids:', common_mut_ids)
# sanity checks
# check if True: should be since these are common
if dr_id.isin(other_id).sum() == other_id.isin(dr_id).sum():
print('PASS: Cross check on no. of common ids')
else:
sys.exit('FAIL: Cross check on no. of common ids failed')
# check if the common are indeed the same!
# bit of a tautology, but better safe than sorry!
common_ids = dr_id[dr_id.isin(other_id)]
common_ids = common_ids.reset_index()
common_ids.columns = ['index', 'id']
common_ids2 = other_id[other_id.isin(dr_id)]
common_ids2 = common_ids2.reset_index()
common_ids2.columns = ['index', 'id2']
# should be True
if common_ids['id'].equals(common_ids2['id2']):
print('PASS: Further cross checks on common ids')
else:
sys.exit('FAIL: Further cross checks on common ids')
# good sanity check: use it later to check gene_sample_counts
expected_gene_samples = (len(meta_gene_dr) + len(meta_gene_other) - common_mut_ids)
print('Expected no. of gene samples:', expected_gene_samples
, '\n=================================================================')
#%% write file
#print(outdir)
out_filename_cid = gene.lower() + '_common_ids.csv'
outfile_cid = outdir + '/' + out_filename_cid
print('Writing file:'
, '\nFile:', outfile_cid
, '\nNo. of rows:', len(common_ids)
, '\n=============================================================')
common_ids.to_csv(outfile_cid, index = False)
del(out_filename_cid)
# clear variables
del(dr_id, other_id, meta_data_dr, meta_data_other, common_ids, common_mut_ids, common_ids2)
#%% Now extract 'all' gene specific nsSNP mutations: i.e 'nssnp_match'
print('Extracting nsSNP match:', gene, 'mutations from cols:\n'
, dr_muts_col, 'and', other_muts_col, 'using string match:'
, '\n===================================================================')
#meta_gene_all = meta_data.loc[meta_data[dr_muts_col].str.contains(gene_match, na = False) | meta_data[other_muts_col].str.contains(gene_match, na = False) ]
meta_gene_all = meta_data.loc[meta_data[dr_muts_col].str.contains(nssnp_match, na = False, regex = True, case = False) | meta_data[other_muts_col].str.contains(nssnp_match, na = False, regex = True, case = False) ]
extracted_gene_samples = meta_gene_all['id'].nunique()
print('RESULT: actual no. of gene samples extracted:', extracted_gene_samples
, '\n===================================================================')
# sanity check: length of gene samples
print('Performing sanity check:')
if extracted_gene_samples == expected_gene_samples:
print('PASS: expected & actual no. of nssnp gene samples match'
, '\nNo. of gene samples:', len(meta_gene_all)
, '\n=========================================================')
else:
sys.exit('FAIL: Length mismatch in gene samples!')
# count NA in drug column
gene_na = meta_gene_all[drug].isna().sum()
print('No. of gene samples without', drug, 'testing:', gene_na)
# use it later to check number of complete samples from LF data
comp_gene_samples = len(meta_gene_all) - gene_na
print('Complete gene samples tested for', drug, ':', comp_gene_samples)
print('=================================================================')
# Comment: This is still dirty data since these
# are samples that have nsSNP muts, but can have others as well
# since the format for mutations is mut1; mut2, etc.
print('This is still dirty data: samples have ', nssnp_match, 'muts but may have others as well'
, '\nsince the format for mutations is mut1; mut2, etc.'
, '\n=============================================================')
print('Performing tidy_split(): to separate the mutations into indivdual rows')
#=========
# DF1: dr_muts_col
#=========
########
# tidy_split(): on dr_muts_col column and remove leading white spaces
########
col_to_split1 = dr_muts_col
print ('Firstly, applying tidy split on dr muts df', meta_gene_dr.shape
, '\ncolumn name to apply tidy_split():', col_to_split1
, '\n============================================================')
# apply tidy_split()
dr_WF0 = tidy_split(meta_gene_dr, col_to_split1, sep = ';')
# remove leading white space else these are counted as distinct mutations as well
dr_WF0[dr_muts_col] = dr_WF0[dr_muts_col].str.lstrip()
# extract only the samples/rows with nssnp_match
#dr_gene_WF0 = dr_WF0.loc[dr_WF0[dr_muts_col].str.contains(gene_match)]
dr_gene_WF0 = dr_WF0.loc[dr_WF0[dr_muts_col].str.contains(nssnp_match, regex = True, case = False)]
print('Lengths after tidy split and extracting', nssnp_match, 'muts:'
, '\nOld length:' , len(meta_gene_dr)
, '\nLength after split:', len(dr_WF0)
, '\nLength of nssnp df:', len(dr_gene_WF0)
, '\nExpected len:', dr_gene_count
, '\n=============================================================')
if len(dr_gene_WF0) == dr_gene_count:
print('PASS: length matches expected length'
, '\n===============================================================')
else:
sys.exit('FAIL: lengths mismatch')
# count the freq of 'dr_muts' samples
dr_muts_df = dr_gene_WF0 [['id', dr_muts_col]]
print('Dim of dr_muts_df:', dr_muts_df.shape)
# add freq column
dr_muts_df['dr_sample_freq'] = dr_muts_df.groupby('id')['id'].transform('count')
#dr_muts_df['dr_sample_freq'] = dr_muts_df.loc[dr_muts_df.groupby('id')].transform('count')
print('Revised dim of dr_muts_df:', dr_muts_df.shape)
c1 = dr_muts_df.dr_sample_freq.value_counts()
print('Counting no. of sample frequency:\n', c1
, '\n===================================================================')
# sanity check: length of gene samples
if len(dr_gene_WF0) == c1.sum():
print('PASS: WF data has expected length'
, '\nLength of dr_gene WFO:', c1.sum()
, '\n===============================================================')
else:
sys.exit('FAIL: length mismatch!')
# Important: Assign 'column name' on which split was performed as an extra column
# This is so you can identify if mutations are dr_type or other in the final df
dr_df = dr_gene_WF0.assign(mutation_info = dr_muts_col)
print('Dim of dr_df:', dr_df.shape
, '\n=============================================================='
, '\nEnd of tidy split() on dr_muts, and added an extra column relecting mut_category'
, '\n===============================================================')
#%%
#=========
# DF2: other_mutations_pyrazinamdie
#=========
########
# tidy_split(): on other_muts_col column and remove leading white spaces
########
col_to_split2 = other_muts_col
print ('applying second tidy split() separately on other muts df', meta_gene_other.shape
, '\ncolumn name to apply tidy_split():', col_to_split2
, '\n============================================================')
# apply tidy_split()
other_WF1 = tidy_split(meta_gene_other, col_to_split2, sep = ';')
# remove the leading white spaces in the column
other_WF1[other_muts_col] = other_WF1[other_muts_col].str.strip()
# extract only the samples/rows with nssnp_match
#other_gene_WF1 = other_WF1.loc[other_WF1[other_muts_col].str.contains(gene_match)]
other_gene_WF1 = other_WF1.loc[other_WF1[other_muts_col].str.contains(nssnp_match, regex = True, case = False)]
print('Lengths after tidy split and extracting', gene_match, 'muts:',
'\nOld length:' , len(meta_gene_other),
'\nLength after split:', len(other_WF1),
'\nLength of nssnp df:', len(other_gene_WF1),
'\nExpected len:', other_gene_count
, '\n=============================================================')
if len(other_gene_WF1) == other_gene_count:
print('PASS: length matches expected length'
, '\n===============================================================')
else:
sys.exit('FAIL: lengths mismatch')
# count the freq of 'other muts' samples
other_muts_df = other_gene_WF1 [['id', other_muts_col]]
print('Dim of other_muts_df:', other_muts_df.shape)
# add freq column
other_muts_df['other_sample_freq'] = other_muts_df.groupby('id')['id'].transform('count')
print('Revised dim of other_muts_df:', other_muts_df.shape)
c2 = other_muts_df.other_sample_freq.value_counts()
print('Counting no. of sample frequency:\n', c2)
print('=================================================================')
# sanity check: length of gene samples
if len(other_gene_WF1) == c2.sum():
print('PASS: WF data has expected length'
, '\nLength of other_gene WFO:', c2.sum()
, '\n===============================================================')
else:
sys.exit('FAIL: Length mismatch')
# Important: Assign 'column name' on which split was performed as an extra column
# This is so you can identify if mutations are dr_type or other in the final df
other_df = other_gene_WF1.assign(mutation_info = other_muts_col)
print('dim of other_df:', other_df.shape
, '\n==============================================================='
, '\nEnd of tidy split() on other_muts, and added an extra column relecting mut_category'
, '\n===============================================================')
#%%
#==========
# Concatentating the two dfs: equivalent of rbind in R
#==========
# Important: Change column names to allow concat:
# dr_muts.. & other_muts : 'mutation'
print('Now concatenating the two dfs by row'
, '\nFirst assigning a common colname: "mutation" to the col containing muts'
, '\nThis is done for both dfs'
, '\n===================================================================')
dr_df.columns
dr_df.rename(columns = {dr_muts_col: 'mutation'}, inplace = True)
dr_df.columns
other_df.columns
other_df.rename(columns = {other_muts_col: 'mutation'}, inplace = True)
other_df.columns
if len(dr_df.columns) == len(other_df.columns):
print('Checking dfs for concatening by rows:'
, '\nDim of dr_df:', dr_df.shape
, '\nDim of other_df:', other_df.shape
, '\nExpected nrows:', len(dr_df) + len(other_df)
, '\n=============================================================')
else:
sys.exit('FAIL: No. of cols mismatch for concatenating')
# checking colnames before concat
print('Checking colnames BEFORE concatenating the two dfs...')
if (set(dr_df.columns) == set(other_df.columns)):
print('PASS: column names match necessary for merging two dfs')
else:
sys.exit('FAIL: Colnames mismatch for concatenating!')
# concatenate (axis = 0): Rbind
print('Now appending the two dfs: Rbind')
gene_LF_comb = pd.concat([dr_df, other_df], ignore_index = True, axis = 0)
print('Finding stop mutations in concatenated df')
stop_muts = gene_LF_comb['mutation'].str.contains('\*').sum()
if stop_muts == 0:
print('PASS: No stop mutations detected')
else:
print('stop mutations detected'
, '\nNo. of stop muts:', stop_muts, '\n'
, gene_LF_comb.groupby(['mutation_info'])['mutation'].apply(lambda x: x[x.str.contains('\*')].count())
, '\nNow removing them')
gene_LF0_nssnp = gene_LF_comb[~gene_LF_comb['mutation'].str.contains('\*')]
print('snps only: by subtracting stop muts:', len(gene_LF0_nssnp))
gene_LF0 = gene_LF_comb[gene_LF_comb['mutation'].str.contains(nssnp_match, case = False)]
print('snps only by direct regex:', len(gene_LF0))
if gene_LF0_nssnp.equals(gene_LF0):
print('PASS: regex for extracting nssnp worked correctly & stop mutations successfully removed'
, '\nUsing the regex extracted df')
else:
sys.exit('FAIL: posssibly regex or no of stop mutations'
, 'Regex being used:', nssnp_match)
#sys.exit()
# checking colnames and length after concat
print('Checking colnames AFTER concatenating the two dfs...')
if (set(dr_df.columns) == set(gene_LF0.columns)):
print('PASS: column names match'
, '\n=============================================================')
else:
sys.exit('FAIL: Colnames mismatch AFTER concatenating')
print('Checking concatenated df')
if len(gene_LF0) == (len(dr_df) + len(other_df))- stop_muts :
print('PASS:length of df after concat match'
, '\n===============================================================')
else:
sys.exit('FAIL: length mismatch')
#%%
###########
# This is hopefully clean data, but just double check
# Filter LF data so that you only have
# mutations corresponding to nssnp_match (case insensitive)
# this will be your list you run OR calcs
###########
print('Length of gene_LF0:', len(gene_LF0)
, '\nThis should be what we need. But just double checking and extracting nsSNP for', gene
, '\nfrom LF0 (concatenated data) using case insensitive regex match:', nssnp_match)
gene_LF1 = gene_LF0[gene_LF0['mutation'].str.contains(nssnp_match, regex = True, case = False)]
if len(gene_LF0) == len(gene_LF1):
print('PASS: length of gene_LF0 and gene_LF1 match',
'\nConfirming extraction and concatenating worked correctly'
, '\n==========================================================')
else:
print('FAIL: BUT NOT FATAL!'
, '\ngene_LF0 and gene_LF1 lengths differ'
, '\nsuggesting error in extraction process'
, ' use gene_LF1 for downstreama analysis'
, '\n=========================================================')
print('Length of dfs pre and post processing...'
, '\ngene WF data (unique samples in each row):', extracted_gene_samples
, '\ngene LF data (unique mutation in each row):', len(gene_LF1)
, '\n=============================================================')
#%% sanity check for extraction
# This ought to pass if nsnsp_match happens right at the beginning of creating 'expected_rows'
print('Verifying whether extraction process worked correctly...')
if len(gene_LF1) == expected_rows:
print('PASS: extraction process performed correctly'
, '\nExpected length:', expected_rows
, '\nGot:', len(gene_LF1)
, '\nRESULT: Total no. of mutant sequences for logo plot:', len(gene_LF1)
, '\n=========================================================')
else:
print('FAIL: extraction process has bugs'
, '\nExpected length:', expected_rows
, '\nGot:', len(gene_LF1)
, '\nDebug please'
, '\n=========================================================')
#%%
print('Performing some more sanity checks...')
# From LF1: useful for OR counts
# no. of unique muts
distinct_muts = gene_LF1.mutation.value_counts()
print('Distinct genomic mutations:', len(distinct_muts))
# no. of samples contributing the unique muts
gene_LF1.id.nunique()
print('No.of samples contributing to distinct genomic muts:', gene_LF1.id.nunique())
# no. of dr and other muts
mut_grouped = gene_LF1.groupby('mutation_info').mutation.nunique()
print('No.of unique dr and other muts:\n', gene_LF1.groupby('mutation_info').mutation.nunique())
# sanity check
if len(distinct_muts) == mut_grouped.sum() :
print('PASS:length of LF1 is as expected'
, '\n===============================================================')
else:
print('FAIL: Mistmatch in count of muts'
, '\nExpected count:', len(distinct_muts)
, '\nActual count:', mut_grouped.sum()
, '\nMuts should be distinct within dr* and other* type'
, '\nInspecting...possibly ambiguous muts'
, '\nNo. of possible ambiguous muts:', mut_grouped.sum() - len(distinct_muts)
, '\n=========================================================')
muts_split = list(gene_LF1.groupby('mutation_info'))
dr_muts = muts_split[0][1].mutation
other_muts = muts_split[1][1].mutation
print('splitting muts by mut_info:', muts_split)
print('no.of dr_muts samples:', len(dr_muts))
print('no. of other_muts samples', len(other_muts))
#%%
# IMPORTANT: The same mutation cannot be classed as a drug AND 'others'
if dr_muts.isin(other_muts).sum() & other_muts.isin(dr_muts).sum() > 0:
print('WARNING: Ambiguous muts detected in dr_ and other_ mutation category'
, '\n===============================================================')
else:
print('PASS: NO ambiguous muts detected'
, '\nMuts are unique to dr_ and other_ mutation class'
, '\n=========================================================')
# inspect dr_muts and other muts
if dr_muts.isin(other_muts).sum() & other_muts.isin(dr_muts).sum() > 0:
print('Finding ambiguous muts...'
, '\n========================================================='
, '\nTotal no. of samples in dr_muts present in other_muts:', dr_muts.isin(other_muts).sum()
, '\nThese are:', dr_muts[dr_muts.isin(other_muts)]
, '\n========================================================='
, '\nTotal no. of samples in other_muts present in dr_muts:', other_muts.isin(dr_muts).sum()
, '\nThese are:\n', other_muts[other_muts.isin(dr_muts)]
, '\n=========================================================')
else:
sys.exit('Error: ambiguous muts present, but extraction failed. Debug!')
print('Counting no. of ambiguous muts...')
if dr_muts[dr_muts.isin(other_muts)].nunique() == other_muts[other_muts.isin(dr_muts)].nunique():
common_muts = dr_muts[dr_muts.isin(other_muts)].value_counts().keys().tolist()
print('Distinct no. of ambigiuous muts detected:'+ str(len(common_muts))
, '\nlist of ambiguous mutations (see below):', *common_muts, sep = '\n')
print('\n===========================================================')
else:
print('Error: ambiguous muts detected, but extraction failed. Debug!'
, '\nNo. of ambiguous muts in dr:'
, len(dr_muts[dr_muts.isin(other_muts)].value_counts().keys().tolist())
, '\nNo. of ambiguous muts in other:'
, len(other_muts[other_muts.isin(dr_muts)].value_counts().keys().tolist())
, '\n=========================================================')
#%% clear variables
del(id_dr, id_other, meta_data, meta_gene_dr, meta_gene_other, mut_grouped, muts_split, other_WF1, other_df, other_muts_df, other_gene_count, gene_LF0, gene_na)
del(c1, c2, col_to_split1, col_to_split2, comp_gene_samples, dr_WF0, dr_df, dr_muts_df, dr_gene_WF0, dr_gene_count, expected_gene_samples, other_gene_WF1)
#%%: write file: ambiguous muts
# uncomment as necessary
#print(outdir)
#dr_muts.to_csv('dr_muts.csv', header = True)
#other_muts.to_csv('other_muts.csv', header = True)
out_filename_ambig_muts = gene.lower() + '_ambiguous_muts.csv'
outfile_ambig_muts = outdir + '/' + out_filename_ambig_muts
print('Writing file: ambiguous muts'
, '\nFilename:', outfile_ambig_muts)
inspect = gene_LF1[gene_LF1['mutation'].isin(common_muts)]
inspect.to_csv(outfile_ambig_muts, index = False)
print('Finished writing:', out_filename_ambig_muts
, '\nNo. of rows:', len(inspect)
, '\nNo. of cols:', len(inspect.columns)
, '\nNo. of rows = no. of samples with the ambiguous muts present:'
, dr_muts.isin(other_muts).sum() + other_muts.isin(dr_muts).sum()
, '\n=============================================================')
del(out_filename_ambig_muts)
#%% end of data extraction and some files writing. Below are some more files writing.
#=============================================================================
#%% Formatting df: read aa dict and pull relevant info
print('Now some more formatting:'
, '\nread aa dict and pull relevant info'
, '\nformat mutations:'
, '\nsplit mutation into mCSM style muts: '
, '\nFormatting mutation in mCSM style format: {WT}<POS>{MUT}'
, '\nassign aa properties: adding 2 cols at a time for each prop'
, '\n===================================================================')
# BEWARE hardcoding : only works as we are adding aa prop once for wt and once for mut
# in each lookup cycle
ncol_mutf_add = 3 # mut split into 3 cols
ncol_aa_add = 2 # 2 aa prop add (wt & mut) in each mapping
#===========
# Split 'mutation' column into three: wild_type, position and
# mutant_type separately. Then map three letter code to one using
# reference_dict imported at the beginning.
# After importing, convert to mutation to lowercase for compatibility with dict
#===========
gene_LF1['mutation'] = gene_LF1.loc[:, 'mutation'].str.lower()
print('wt regex being used:', wt_regex
, '\nmut regex being used:', mut_regex
, '\nposition regex being used:', pos_regex)
mylen0 = len(gene_LF1.columns)
#=======
# Iterate through the dict, create a lookup dict i.e
# lookup_dict = {three_letter_code: one_letter_code}.
# lookup dict should be the key and the value (you want to create a column for)
# Then use this to perform the mapping separetly for wild type and mutant cols.
# The three letter code is extracted using a string match match from the dataframe and then converted
# to 'pandas series'since map only works in pandas series
#=======
print('Adding', ncol_mutf_add, 'more cols:\n')
# initialise a sub dict that is lookup dict for three letter code to 1-letter code
# adding three more cols
lookup_dict = dict()
for k, v in my_aa_dict.items():
lookup_dict[k] = v['one_letter_code']
#wt = gene_LF1['mutation'].str.extract(gene_regex).squeeze()converts to a series that map works on
wt = gene_LF1['mutation'].str.extract(wt_regex).squeeze()
gene_LF1['wild_type'] = wt.map(lookup_dict)
#mut = gene_LF1['mutation'].str.extract('\d+(\w{3})$').squeeze()
mut = gene_LF1['mutation'].str.extract(mut_regex).squeeze()
gene_LF1['mutant_type'] = mut.map(lookup_dict)
# extract position info from mutation column separetly using string match
#gene_LF1['position'] = gene_LF1['mutation'].str.extract(r'(\d+)')
gene_LF1['position'] = gene_LF1['mutation'].str.extract(pos_regex)
mylen1 = len(gene_LF1.columns)
# sanity checks
print('checking if 3-letter wt&mut residue extraction worked correctly')
if wt.isna().sum() & mut.isna().sum() == 0:
print('PASS: 3-letter wt&mut residue extraction worked correctly:'
, '\nNo NAs detected:'
, '\nwild-type\n', wt
, '\nmutant-type\n', mut
, '\ndim of df:', gene_LF1.shape)
else:
print('FAIL: 3-letter wt&mut residue extraction failed'
, '\nNo NAs detected:'
, '\nwild-type\n', wt
, '\nmutant-type\n', mut
, '\ndim of df:', gene_LF1.shape)
if mylen1 == mylen0 + ncol_mutf_add:
print('PASS: successfully added', ncol_mutf_add, 'cols'
, '\nold length:', mylen0
, '\nnew len:', mylen1)
else:
print('FAIL: failed to add cols:'
, '\nold length:', mylen0
, '\nnew len:', mylen1)
# clear variables
del(k, v, wt, mut, lookup_dict)
#=========
# iterate through the dict, create a lookup dict that i.e
# lookup_dict = {three_letter_code: aa_prop_water}
# Do this for both wild_type and mutant as above.
#=========
print('Adding', ncol_aa_add, 'more cols:\n')
# initialise a sub dict that is lookup dict for three letter code to aa prop
# adding two more cols
lookup_dict = dict()
for k, v in my_aa_dict.items():
lookup_dict[k] = v['aa_prop_water']
#print(lookup_dict)
wt = gene_LF1['mutation'].str.extract(wt_regex).squeeze()
gene_LF1['wt_prop_water'] = wt.map(lookup_dict)
#mut = gene_LF1['mutation'].str.extract('\d+(\w{3})$').squeeze()
mut = gene_LF1['mutation'].str.extract(mut_regex).squeeze()
gene_LF1['mut_prop_water'] = mut.map(lookup_dict)
mylen2 = len(gene_LF1.columns)
# sanity checks
print('checking if 3-letter wt&mut residue extraction worked correctly')
if wt.isna().sum() & mut.isna().sum() == 0:
print('PASS: 3-letter wt&mut residue extraction worked correctly:'
, '\nNo NAs detected:'
, '\nwild-type\n', wt
, '\nmutant-type\n', mut
, '\ndim of df:', gene_LF1.shape)
else:
print('FAIL: 3-letter wt&mut residue extraction failed'
, '\nNo NAs detected:'
, '\nwild-type\n', wt
, '\nmutant-type\n', mut
, '\ndim of df:', gene_LF1.shape)
if mylen2 == mylen1 + ncol_aa_add:
print('PASS: successfully added', ncol_aa_add, 'cols'
, '\nold length:', mylen1
, '\nnew len:', mylen2)
else:
print('FAIL: failed to add cols:'
, '\nold length:', mylen1
, '\nnew len:', mylen2)
# clear variables
del(k, v, wt, mut, lookup_dict)
#========
# iterate through the dict, create a lookup dict that i.e
# lookup_dict = {three_letter_code: aa_prop_polarity}
# Do this for both wild_type and mutant as above.
#=========
print('Adding', ncol_aa_add, 'more cols:\n')
# initialise a sub dict that is lookup dict for three letter code to aa prop
# adding two more cols
lookup_dict = dict()
for k, v in my_aa_dict.items():
lookup_dict[k] = v['aa_prop_polarity']
#print(lookup_dict)
wt = gene_LF1['mutation'].str.extract(wt_regex).squeeze()
gene_LF1['wt_prop_polarity'] = wt.map(lookup_dict)
#mut = gene_LF1['mutation'].str.extract(r'\d+(\w{3})$').squeeze()
mut = gene_LF1['mutation'].str.extract(mut_regex).squeeze()
gene_LF1['mut_prop_polarity'] = mut.map(lookup_dict)
mylen3 = len(gene_LF1.columns)
# sanity checks
print('checking if 3-letter wt&mut residue extraction worked correctly')
if wt.isna().sum() & mut.isna().sum() == 0:
print('PASS: 3-letter wt&mut residue extraction worked correctly:'
, '\nNo NAs detected:'
, '\nwild-type\n', wt
, '\nmutant-type\n', mut
, '\ndim of df:', gene_LF1.shape)
else:
print('FAIL: 3-letter wt&mut residue extraction failed'
, '\nNo NAs detected:'
, '\nwild-type\n', wt
, '\nmutant-type\n', mut
, '\ndim of df:', gene_LF1.shape)
if mylen3 == mylen2 + ncol_aa_add:
print('PASS: successfully added', ncol_aa_add, 'cols'
, '\nold length:', mylen1
, '\nnew len:', mylen2)
else:
print('FAIL: failed to add cols:'
, '\nold length:', mylen1
, '\nnew len:', mylen2)
# clear variables
del(k, v, wt, mut, lookup_dict)
#========
# iterate through the dict, create a lookup dict that i.e
# lookup_dict = {three_letter_code: aa_calcprop}
# Do this for both wild_type and mutant as above.
#=========
print('Adding', ncol_aa_add, 'more cols:\n')
lookup_dict = dict()
for k, v in my_aa_dict.items():
lookup_dict[k] = v['aa_calcprop']
#print(lookup_dict)
wt = gene_LF1['mutation'].str.extract(wt_regex).squeeze()
gene_LF1['wt_calcprop'] = wt.map(lookup_dict)
mut = gene_LF1['mutation'].str.extract(mut_regex).squeeze()
gene_LF1['mut_calcprop'] = mut.map(lookup_dict)
mylen4 = len(gene_LF1.columns)
# sanity checks
print('checking if 3-letter wt&mut residue extraction worked correctly')
if wt.isna().sum() & mut.isna().sum() == 0:
print('PASS: 3-letter wt&mut residue extraction worked correctly:'
, '\nNo NAs detected:'
, '\nwild-type\n', wt
, '\nmutant-type\n', mut
, '\ndim of df:', gene_LF1.shape)
else:
print('FAIL: 3-letter wt&mut residue extraction failed'
, '\nNo NAs detected:'
, '\nwild-type\n', wt
, '\nmutant-type\n', mut
, '\ndim of df:', gene_LF1.shape)
if mylen4 == mylen3 + ncol_aa_add:
print('PASS: successfully added', ncol_aa_add, 'cols'
, '\nold length:', mylen3
, '\nnew len:', mylen4)
else:
print('FAIL: failed to add cols:'
, '\nold length:', mylen3
, '\nnew len:', mylen4)
# clear variables
del(k, v, wt, mut, lookup_dict)
########
# combine the wild_type+poistion+mutant_type columns to generate
# mutationinformation (matches mCSM output field)
# Remember to use .map(str) for int col types to allow string concatenation
#########
gene_LF1['mutationinformation'] = gene_LF1['wild_type'] + gene_LF1.position.map(str) + gene_LF1['mutant_type']
print('Created column: mutationinformation'
, '\n=====================================================================\n'
, gene_LF1.mutationinformation.head(10))
#%% Write file: mCSM muts
snps_only = pd.DataFrame(gene_LF1['mutationinformation'].unique())
snps_only.head()
# assign column name
snps_only.columns = ['mutationinformation']
# count how many positions this corresponds to
pos_only = pd.DataFrame(gene_LF1['position'].unique())
print('Checking NA in snps...')# should be 0
if snps_only.mutationinformation.isna().sum() == 0:
print ('PASS: NO NAs/missing entries for SNPs'
, '\n===============================================================')
else:
sys.exit('FAIL: SNP has NA, Possible mapping issues from dict?')
out_filename_mcsmsnps = gene.lower() + '_mcsm_snps.csv'
outfile_mcsmsnps = outdir + '/' + out_filename_mcsmsnps
print('Writing file: mCSM style muts'
, '\nFile:', outfile_mcsmsnps
, '\nmutation format (SNP): {WT}<POS>{MUT}'
, '\nNo. of distinct muts:', len(snps_only)
, '\nNo. of distinct positions:', len(pos_only)
, '\n=============================================================')
snps_only.to_csv(outfile_mcsmsnps, header = False, index = False)
print('Finished writing:', outfile_mcsmsnps
, '\nNo. of rows:', len(snps_only)
, '\nNo. of cols:', len(snps_only.columns)
, '\n=============================================================')
del(out_filename_mcsmsnps)
#%% Write file: gene_metadata (i.e gene_LF1)
# where each row has UNIQUE mutations NOT unique sample ids
out_filename_metadata = gene.lower() + '_metadata.csv'
outfile_metadata = outdir + '/' + out_filename_metadata
print('Writing file: LF formatted data'
, '\nFile:', outfile_metadata
, '\n============================================================')
gene_LF1.to_csv(outfile_metadata, header = True, index = False)
print('Finished writing:', outfile_metadata
, '\nNo. of rows:', len(gene_LF1)
, '\nNo. of cols:', len(gene_LF1.columns)
, '\n=============================================================')
del(out_filename_metadata)
#%% write file: mCSM style but with repitions for MSA and logo plots
all_muts_msa = pd.DataFrame(gene_LF1['mutationinformation'])
all_muts_msa.head()
# assign column name
all_muts_msa.columns = ['mutationinformation']
# make sure it is string
all_muts_msa.columns.dtype
# sort the column
all_muts_msa_sorted = all_muts_msa.sort_values(by = 'mutationinformation')
# create an extra column with protein name
#all_muts_msa_sorted = all_muts_msa_sorted.assign(fasta_name = '3PL1')
#all_muts_msa_sorted.head()
# rearrange columns so the fasta name is the first column (required for mutate.script)
#all_muts_msa_sorted = all_muts_msa_sorted[['fasta_name', 'mutationinformation']]
all_muts_msa_sorted.head()
print('Checking NA in snps...')# should be 0
if all_muts_msa.mutationinformation.isna().sum() == 0:
print ('PASS: NO NAs/missing entries for SNPs'
, '\n===============================================================')
else:
sys.exit('FAIL: SNP has NA, Possible mapping issues from dict?')
out_filename_msa = gene.lower() +'_all_muts_msa.csv'
outfile_msa = outdir + '/' + out_filename_msa
print('Writing file: mCSM style muts for msa',
'\nFile:', outfile_msa,
'\nmutation format (SNP): {WT}<POS>{MUT}',
'\nNo.of lines of msa:', len(all_muts_msa))
all_muts_msa_sorted.to_csv(outfile_msa, header = False, index = False)
print('Finished writing:', outfile_msa
, '\nNo. of rows:', len(all_muts_msa)
, '\nNo. of cols:', len(all_muts_msa.columns)
, '\n=============================================================')
del(out_filename_msa)
#%% write file for mutational positions
# count how many positions this corresponds to
pos_only = pd.DataFrame(gene_LF1['position'].unique())
# assign column name
pos_only.columns = ['position']
# make sure dtype of column position is int or numeric and not string
pos_only.position.dtype
pos_only['position'] = pos_only['position'].astype(int)
pos_only.position.dtype
# sort by position value
pos_only_sorted = pos_only.sort_values(by = 'position', ascending = True)
out_filename_pos = gene.lower() + '_mutational_positons.csv'
outfile_pos = outdir + '/' + out_filename_pos
print('Writing file: mutational positions'
, '\nFile:', outfile_pos
, '\nNo. of distinct positions:', len(pos_only_sorted)
, '\n=============================================================')
pos_only_sorted.to_csv(outfile_pos, header = True, index = False)
print('Finished writing:', outfile_pos
, '\nNo. of rows:', len(pos_only_sorted)
, '\nNo. of cols:', len(pos_only_sorted.columns)
, '\n=============================================================')
del(out_filename_pos)
#=======================================================================
print(u'\u2698' * 50,
'\nEnd of script: Data extraction and writing files'
'\n' + u'\u2698' * 50 )
#%% end of script