Mercurial > repos > jackcurragh > trips_viz_bam_to_sqlite
view trips_bam_to_sqlite_builtin/bam_to_sqlite.py @ 13:df4bb52b226d draft
Uploaded
author | jackcurragh |
---|---|
date | Thu, 03 Nov 2022 12:25:58 +0000 |
parents | |
children |
line wrap: on
line source
import sys import pysam import operator import os import time import sqlite3 from sqlitedict import SqliteDict def tran_to_genome(tran, pos, transcriptome_info_dict): #print ("tran",list(transcriptome_info_dict)) traninfo = transcriptome_info_dict[tran] chrom = traninfo["chrom"] strand = traninfo["strand"] exons = sorted(traninfo["exons"]) #print exons if strand == "+": exon_start = 0 for tup in exons: exon_start = tup[0] exonlen = tup[1] - tup[0] if pos > exonlen: pos = (pos - exonlen)-1 else: break genomic_pos = (exon_start+pos)-1 elif strand == "-": exon_start = 0 for tup in exons[::-1]: exon_start = tup[1] exonlen = tup[1] - tup[0] if pos > exonlen: pos = (pos - exonlen)-1 else: break genomic_pos = (exon_start-pos)+1 return (chrom, genomic_pos) # Takes a dictionary with a readname as key and a list of lists as value, each sub list has consists of two elements a transcript and the position the read aligns to in the transcript # This function will count the number of genes that the transcripts correspond to and if less than or equal to 3 will add the relevant value to transcript_counts_dict def processor(process_chunk, master_read_dict, transcriptome_info_dict,master_dict,readseq, unambig_read_length_dict): readlen = len(readseq) ambiguously_mapped_reads = 0 #get the read name read = list(process_chunk)[0] read_list = process_chunk[read] # a list of lists of all transcripts the read aligns to and the positions #used to store different genomic poistions genomic_positions = [] #This section is just to get the different genomic positions the read aligns to for listname in process_chunk[read]: tran = listname[0].replace("-","_").replace("(","").replace(")","") pos = int(listname[1]) genomic_pos = tran_to_genome(tran, pos, transcriptome_info_dict) #print ("genomic pos",genomic_pos) if genomic_pos not in genomic_positions: genomic_positions.append(genomic_pos) #If the read maps unambiguously if len(genomic_positions) == 1: if readlen not in unambig_read_length_dict: unambig_read_length_dict[readlen] = 0 unambig_read_length_dict[readlen] += 1 #assume this read aligns to a noncoding position, if we find that it does align to a coding region change this to True coding=False # For each transcript this read alings to for listname in process_chunk[read]: #get the transcript name tran = listname[0].replace("-","_").replace("(","").replace(")","") #If we haven't come across this transcript already then add to master_read_dict if tran not in master_read_dict: master_read_dict[tran] = {"ambig":{}, "unambig":{}, "mismatches":{}, "seq":{}} #get the raw unedited positon, and read tags pos = int(listname[1]) read_tags = listname[2] #If there is mismatches in this line, then modify the postion and readlen (if mismatches at start or end) and add mismatches to dictionary nm_tag = 0 for tag in read_tags: if tag[0] == "NM": nm_tag = int(tag[1]) if nm_tag > 0: md_tag = "" for tag in read_tags: if tag[0] == "MD": md_tag = tag[1] pos_modifier, readlen_modifier,mismatches = get_mismatch_pos(md_tag,pos,readlen,master_read_dict,tran,readseq) # Count the mismatches (we only do this for unambiguous) for mismatch in mismatches: #Ignore mismatches appearing in the first position (due to non templated addition) if mismatch != 0: char = mismatches[mismatch] mismatch_pos = pos + mismatch if mismatch_pos not in master_read_dict[tran]["seq"]: master_read_dict[tran]["seq"][mismatch_pos] = {} if char not in master_read_dict[tran]["seq"][mismatch_pos]: master_read_dict[tran]["seq"][mismatch_pos][char] = 0 master_read_dict[tran]["seq"][mismatch_pos][char] += 1 # apply the modifiers #pos = pos+pos_modifier #readlen = readlen - readlen_modifier try: cds_start = transcriptome_info_dict[tran]["cds_start"] cds_stop = transcriptome_info_dict[tran]["cds_stop"] if pos >= cds_start and pos <= cds_stop: coding=True except: pass if readlen in master_read_dict[tran]["unambig"]: if pos in master_read_dict[tran]["unambig"][readlen]: master_read_dict[tran]["unambig"][readlen][pos] += 1 else: master_read_dict[tran]["unambig"][readlen][pos] = 1 else: master_read_dict[tran]["unambig"][readlen] = {pos:1} if coding == True: master_dict["unambiguous_coding_count"] += 1 elif coding == False: master_dict["unambiguous_non_coding_count"] += 1 else: ambiguously_mapped_reads += 1 for listname in process_chunk[read]: tran = listname[0].replace("-","_").replace("(","").replace(")","") if tran not in master_read_dict: master_read_dict[tran] = {"ambig":{}, "unambig":{}, "mismatches":{}, "seq":{}} pos = int(listname[1]) read_tags = listname[2] nm_tag = 0 for tag in read_tags: if tag[0] == "NM": nm_tag = int(tag[1]) if nm_tag > 0: md_tag = "" for tag in read_tags: if tag[0] == "MD": md_tag = tag[1] pos_modifier, readlen_modifier,mismatches = get_mismatch_pos(md_tag,pos,readlen,master_read_dict,tran,readseq) # apply the modifiers #pos = pos+pos_modifier #readlen = readlen - readlen_modifier if readlen in master_read_dict[tran]["ambig"]: if pos in master_read_dict[tran]["ambig"][readlen]: master_read_dict[tran]["ambig"][readlen][pos] += 1 else: master_read_dict[tran]["ambig"][readlen][pos] = 1 else: master_read_dict[tran]["ambig"][readlen] = {pos:1} return ambiguously_mapped_reads def get_mismatch_pos(md_tag,pos,readlen,master_read_dict,tran,readseq): nucs = ["A","T","G","C"] mismatches = {} total_so_far = 0 prev_char = "" for char in md_tag: if char in nucs: if prev_char != "": total_so_far += int(prev_char) prev_char = "" mismatches[total_so_far+len(mismatches)] = (readseq[total_so_far+len(mismatches)]) else: if char != "^" and char != "N": if prev_char == "": prev_char = char else: total_so_far += int(prev_char+char) prev_char = "" readlen_modifier = 0 pos_modifier = 0 five_ok = False three_ok = False while five_ok == False: for i in range(0,readlen): if i in mismatches: pos_modifier += 1 readlen_modifier += 1 else: five_ok = True break five_ok = True while three_ok == False: for i in range(readlen-1,0,-1): if i in mismatches: readlen_modifier += 1 else: three_ok = True break three_ok = True return (pos_modifier, readlen_modifier, mismatches) def process_bam(bam_filepath, transcriptome_info_dict_path,outputfile): desc = "NULL" start_time = time.time() study_dict ={} nuc_count_dict = {"mapped":{},"unmapped":{}} dinuc_count_dict = {} threeprime_nuc_count_dict = {"mapped":{},"unmapped":{}} read_length_dict = {} unambig_read_length_dict = {} unmapped_dict = {} master_dict = {"unambiguous_non_coding_count":0,"unambiguous_coding_count":0,"current_dir":os.getcwd()} transcriptome_info_dict = {} connection = sqlite3.connect(transcriptome_info_dict_path) cursor = connection.cursor() cursor.execute("SELECT transcript,cds_start,cds_stop,length,strand,chrom,tran_type from transcripts;") result = cursor.fetchall() for row in result: transcriptome_info_dict[str(row[0])] = {"cds_start":row[1],"cds_stop":row[2],"length":row[3],"strand":row[4],"chrom":row[5],"exons":[],"tran_type":row[6]} #print list(transcriptome_info_dict)[:10] cursor.execute("SELECT * from exons;") result = cursor.fetchall() for row in result: transcriptome_info_dict[str(row[0])]["exons"].append((row[1],row[2])) #it might be the case that there are no multimappers, so set this to 0 first to avoid an error, it will be overwritten later if there is multimappers multimappers = 0 unmapped_reads = 0 unambiguous_coding_count = 0 unambiguous_non_coding_count = 0 trip_periodicity_reads = 0 final_offsets = {"fiveprime":{"offsets":{}, "read_scores":{}}, "threeprime":{"offsets":{}, "read_scores":{}}} master_read_dict = {} prev_seq = "" process_chunk = {"read_name":[["placeholder_tran","1","28"]]} mapped_reads = 0 ambiguously_mapped_reads = 0 master_trip_dict = {"fiveprime":{}, "threeprime":{}} master_offset_dict = {"fiveprime":{}, "threeprime":{}} master_metagene_stop_dict = {"fiveprime":{}, "threeprime":{}} os.system(f'samtools sort -n {bam_filepath} -o {bam_filepath}_n_sorted.bam') pysam.set_verbosity(0) infile = pysam.Samfile(f"{bam_filepath}_n_sorted.bam", "rb") header = infile.header["HD"] unsorted = False if "SO" in header: print("Sorting order: "+header["SO"]) if header["SO"] != "queryname": print("Sorting order is not queryname") unsorted = True else: unsorted = True if unsorted == True: print ("ERROR: Bam file appears to be unsorted or not sorted by read name. To sort by read name use the command: samtools sort -n input.bam output.bam") print (header,bam_filepath) sys.exit() total_bam_lines = 0 all_ref_ids = infile.references for read in infile.fetch(until_eof=True): total_bam_lines += 1 if not read.is_unmapped: ref = read.reference_id tran = (all_ref_ids[ref]).split(".")[0] mapped_reads += 1 if mapped_reads%1000000 == 0: print ("{} reads parsed at {}".format(mapped_reads,(time.time()-start_time))) pos = read.reference_start readname = read.query_name read_tags = read.tags if readname == list(process_chunk)[0]: process_chunk[readname].append([tran,pos,read_tags]) #if the current read is different from previous reads send 'process_chunk' to the 'processor' function, then start 'process_chunk' over using current read else: if list(process_chunk)[0] != "read_name": #At this point we work out readseq, we do this for multiple reasons, firstly so we don't count the sequence from a read multiple times, just because # it aligns multiple times and secondly we only call read.seq once (read.seq is computationally expensive) seq = read.seq readlen = len(seq) # Note if a read maps ambiguously it will still be counted toward the read length distribution (however it will only be counted once, not each time it maps) if readlen not in read_length_dict: read_length_dict[readlen] = 0 read_length_dict[readlen] += 1 if readlen not in nuc_count_dict["mapped"]: nuc_count_dict["mapped"][readlen] = {} if readlen not in threeprime_nuc_count_dict["mapped"]: threeprime_nuc_count_dict["mapped"][readlen] = {} if readlen not in dinuc_count_dict: dinuc_count_dict[readlen] = {"AA":0, "TA":0, "GA":0, "CA":0, "AT":0, "TT":0, "GT":0, "CT":0, "AG":0, "TG":0, "GG":0, "CG":0, "AC":0, "TC":0, "GC":0, "CC":0} for i in range(0,len(seq)): if i not in nuc_count_dict["mapped"][readlen]: nuc_count_dict["mapped"][readlen][i] = {"A":0, "T":0, "G":0, "C":0, "N":0} nuc_count_dict["mapped"][readlen][i][seq[i]] += 1 for i in range(0,len(seq)): try: dinuc_count_dict[readlen][seq[i:i+2]] += 1 except: pass for i in range(len(seq),0,-1): dist = i-len(seq) if dist not in threeprime_nuc_count_dict["mapped"][readlen]: threeprime_nuc_count_dict["mapped"][readlen][dist] = {"A":0, "T":0, "G":0, "C":0, "N":0} threeprime_nuc_count_dict["mapped"][readlen][dist][seq[dist]] += 1 ambiguously_mapped_reads += processor(process_chunk, master_read_dict, transcriptome_info_dict,master_dict,prev_seq, unambig_read_length_dict) process_chunk = {readname:[[tran, pos, read_tags]]} prev_seq = read.seq else: unmapped_reads += 1 # Add this unmapped read to unmapped_dict so we can see what the most frequent unmapped read is. seq = read.seq readlen = len(seq) if seq in unmapped_dict: unmapped_dict[seq] += 1 else: unmapped_dict[seq] = 1 # Populate the nuc_count_dict with this unmapped read if readlen not in nuc_count_dict["unmapped"]: nuc_count_dict["unmapped"][readlen] = {} for i in range(0,len(seq)): if i not in nuc_count_dict["unmapped"][readlen]: nuc_count_dict["unmapped"][readlen][i] = {"A":0, "T":0, "G":0, "C":0, "N":0} nuc_count_dict["unmapped"][readlen][i][seq[i]] += 1 if readlen not in threeprime_nuc_count_dict["unmapped"]: threeprime_nuc_count_dict["unmapped"][readlen] = {} for i in range(len(seq),0,-1): dist = i-len(seq) if dist not in threeprime_nuc_count_dict["unmapped"][readlen]: threeprime_nuc_count_dict["unmapped"][readlen][dist] = {"A":0, "T":0, "G":0, "C":0, "N":0} threeprime_nuc_count_dict["unmapped"][readlen][dist][seq[dist]] += 1 #add stats about mapped/unmapped reads to file dict which will be used for the final report master_dict["total_bam_lines"] = total_bam_lines master_dict["mapped_reads"] = mapped_reads master_dict["unmapped_reads"] = unmapped_reads master_dict["ambiguously_mapped_reads"] = ambiguously_mapped_reads if "read_name" in master_read_dict: del master_read_dict["read_name"] print ("BAM file processed") print ("Creating metagenes, triplet periodicity plots, etc.") for tran in master_read_dict: try: cds_start = int(0 if transcriptome_info_dict[tran]["cds_start"] is None else transcriptome_info_dict[tran]["cds_start"]) cds_stop = int(0 if transcriptome_info_dict[tran]["cds_stop"] is None else transcriptome_info_dict[tran]["cds_stop"]) # print(tran, type(cds_start)) except: print("Exception: ", tran) continue tranlen = transcriptome_info_dict[tran]["length"] #Use this to discard transcripts with no 5' leader or 3' trailer if cds_start > 1 and cds_stop < tranlen and transcriptome_info_dict[tran]["tran_type"] == 1: for primetype in ["fiveprime", "threeprime"]: # Create the triplet periodicity and metainfo plots based on both the 5' and 3' ends of reads for readlength in master_read_dict[tran]["unambig"]: #print "readlength", readlength # for each fiveprime postion for this readlength within this transcript for raw_pos in master_read_dict[tran]["unambig"][readlength]: #print "raw pos", raw_pos trip_periodicity_reads += 1 if primetype == "fiveprime": # get the five prime postion minus the cds start postion real_pos = raw_pos-cds_start rel_stop_pos = raw_pos-cds_stop elif primetype == "threeprime": real_pos = (raw_pos+readlength)-cds_start rel_stop_pos = (raw_pos+readlength)-cds_stop #get the readcount at the raw postion readcount = master_read_dict[tran]["unambig"][readlength][raw_pos] #print "readcount", readcount frame = (real_pos%3) if real_pos >= cds_start and real_pos <= cds_stop: if readlength in master_trip_dict[primetype]: master_trip_dict[primetype][readlength][str(frame)] += readcount else: master_trip_dict[primetype][readlength]= {"0":0.0,"1":0.0,"2":0.0} master_trip_dict[primetype][readlength][str(frame)] += readcount # now populate offset dict with the 'real_positions' upstream of cds_start, these will be used for metainfo dict if real_pos > (-600) and real_pos < (601): if readlength in master_offset_dict[primetype]: if real_pos in master_offset_dict[primetype][readlength]: #print "real pos in offset dict" master_offset_dict[primetype][readlength][real_pos] += readcount else: #print "real pos not in offset dict" master_offset_dict[primetype][readlength][real_pos] = readcount else: #initiliase with zero to avoid missing neighbours below #print "initialising with zeros" master_offset_dict[primetype][readlength]= {} for i in range(-600,601): master_offset_dict[primetype][readlength][i] = 0 master_offset_dict[primetype][readlength][real_pos] += readcount # now populate offset dict with the 'real_positions' upstream of cds_start, these will be used for metainfo dict if rel_stop_pos > (-600) and rel_stop_pos < (601): if readlength in master_metagene_stop_dict[primetype]: if rel_stop_pos in master_metagene_stop_dict[primetype][readlength]: master_metagene_stop_dict[primetype][readlength][rel_stop_pos] += readcount else: master_metagene_stop_dict[primetype][readlength][rel_stop_pos] = readcount else: #initiliase with zero to avoid missing neighbours below master_metagene_stop_dict[primetype][readlength] = {} for i in range(-600,601): master_metagene_stop_dict[primetype][readlength][i] = 0 master_metagene_stop_dict[primetype][readlength][rel_stop_pos] += readcount # master trip dict is now made up of readlengths with 3 frames and a count associated with each frame # create a 'score' for each readlength by putting the max frame count over the second highest frame count for primetype in ["fiveprime", "threeprime"]: for subreadlength in master_trip_dict[primetype]: maxcount = 0 secondmaxcount = 0 for frame in master_trip_dict[primetype][subreadlength]: if master_trip_dict[primetype][subreadlength][frame] > maxcount: maxcount = master_trip_dict[primetype][subreadlength][frame] for frame in master_trip_dict[primetype][subreadlength]: if master_trip_dict[primetype][subreadlength][frame] > secondmaxcount and master_trip_dict[primetype][subreadlength][frame] != maxcount: secondmaxcount = master_trip_dict[primetype][subreadlength][frame] # a perfect score would be 0 meaning there is only a single peak, the worst score would be 1 meaning two highest peaks are the same height master_trip_dict[primetype][subreadlength]["score"] = float(secondmaxcount)/float(maxcount) #This part is to determine what offsets to give each read length print ("Calculating offsets") for primetype in ["fiveprime", "threeprime"]: for readlen in master_offset_dict[primetype]: accepted_len = False max_relative_pos = 0 max_relative_count = 0 for relative_pos in master_offset_dict[primetype][readlen]: # This line is to ensure we don't choose an offset greater than the readlength (in cases of a large peak far up/downstream) if abs(relative_pos) < 10 or abs(relative_pos) > (readlen-10): continue if master_offset_dict[primetype][readlen][relative_pos] > max_relative_count: max_relative_pos = relative_pos max_relative_count = master_offset_dict[primetype][readlen][relative_pos] #print "for readlen {} the max_relative pos is {}".format(readlen, max_relative_pos) if primetype == "fiveprime": # -3 to get from p-site to a-site, +1 to account for 1 based co-ordinates, resulting in -2 overall final_offsets[primetype]["offsets"][readlen] = abs(max_relative_pos-2) elif primetype == "threeprime": # +3 to get from p-site to a-site, -1 to account for 1 based co-ordinates, resulting in +2 overall final_offsets[primetype]["offsets"][readlen] = (max_relative_pos*(-1))+2 #If there are no reads in CDS regions for a specific length, it may not be present in master_trip_dict if readlen in master_trip_dict[primetype]: final_offsets[primetype]["read_scores"][readlen] = master_trip_dict[primetype][readlen]["score"] else: final_offsets[primetype]["read_scores"][readlen] = 0.0 master_read_dict["unmapped_reads"] = unmapped_reads master_read_dict["offsets"] = final_offsets master_read_dict["trip_periodicity"] = master_trip_dict master_read_dict["desc"] = "Null" master_read_dict["mapped_reads"] = mapped_reads master_read_dict["nuc_counts"] = nuc_count_dict master_read_dict["dinuc_counts"] = dinuc_count_dict master_read_dict["threeprime_nuc_counts"] = threeprime_nuc_count_dict master_read_dict["metagene_counts"] = master_offset_dict master_read_dict["stop_metagene_counts"] = master_metagene_stop_dict master_read_dict["read_lengths"] = read_length_dict master_read_dict["unambig_read_lengths"] = unambig_read_length_dict master_read_dict["coding_counts"] = master_dict["unambiguous_coding_count"] master_read_dict["noncoding_counts"] = master_dict["unambiguous_non_coding_count"] master_read_dict["ambiguous_counts"] = master_dict["ambiguously_mapped_reads"] master_read_dict["frequent_unmapped_reads"] = (sorted(unmapped_dict.items(), key=operator.itemgetter(1)))[-2000:] master_read_dict["cutadapt_removed"] = 0 master_read_dict["rrna_removed"] = 0 #If no reads are removed by minus m there won't be an entry in the log file, so initiliase with 0 first and change if there is a line master_read_dict["removed_minus_m"] = 0 master_dict["removed_minus_m"] = 0 # We work out the total counts for 5', cds 3' for differential translation here, would be better to do thisn in processor but need the offsets master_read_dict["unambiguous_all_totals"] = {} master_read_dict["unambiguous_fiveprime_totals"] = {} master_read_dict["unambiguous_cds_totals"] = {} master_read_dict["unambiguous_threeprime_totals"] = {} master_read_dict["ambiguous_all_totals"] = {} master_read_dict["ambiguous_fiveprime_totals"] = {} master_read_dict["ambiguous_cds_totals"] = {} master_read_dict["ambiguous_threeprime_totals"] = {} print ("calculating transcript counts") for tran in master_read_dict: if tran in transcriptome_info_dict: five_total = 0 cds_total = 0 three_total = 0 ambig_five_total = 0 ambig_cds_total = 0 ambig_three_total = 0 cds_start = transcriptome_info_dict[tran]["cds_start"] cds_stop = transcriptome_info_dict[tran]["cds_stop"] for readlen in master_read_dict[tran]["unambig"]: if readlen in final_offsets["fiveprime"]["offsets"]: offset = final_offsets["fiveprime"]["offsets"][readlen] else: offset = 15 for pos in master_read_dict[tran]["unambig"][readlen]: real_pos = pos+offset if cds_start is None or cds_stop is None: three_total += master_read_dict[tran]["unambig"][readlen][pos] else: if real_pos <cds_start: five_total += master_read_dict[tran]["unambig"][readlen][pos] elif real_pos >=cds_start and real_pos <= cds_stop: cds_total += master_read_dict[tran]["unambig"][readlen][pos] elif real_pos > cds_stop: three_total += master_read_dict[tran]["unambig"][readlen][pos] master_read_dict["unambiguous_all_totals"][tran] = five_total+cds_total+three_total master_read_dict["unambiguous_fiveprime_totals"][tran] = five_total master_read_dict["unambiguous_cds_totals"][tran] = cds_total master_read_dict["unambiguous_threeprime_totals"][tran] = three_total for readlen in master_read_dict[tran]["ambig"]: if readlen in final_offsets["fiveprime"]["offsets"]: offset = final_offsets["fiveprime"]["offsets"][readlen] else: offset = 15 for pos in master_read_dict[tran]["ambig"][readlen]: if cds_start is None or cds_stop is None: ambig_three_total += master_read_dict[tran]["ambig"][readlen][pos] else: real_pos = pos+offset if real_pos < cds_start: ambig_five_total += master_read_dict[tran]["ambig"][readlen][pos] elif real_pos >=cds_start and real_pos <= cds_stop: ambig_cds_total += master_read_dict[tran]["ambig"][readlen][pos] elif real_pos > cds_stop: ambig_three_total += master_read_dict[tran]["ambig"][readlen][pos] master_read_dict["ambiguous_all_totals"][tran] = five_total+cds_total+three_total+ambig_five_total+ambig_cds_total+ambig_three_total master_read_dict["ambiguous_fiveprime_totals"][tran] = five_total+ambig_five_total master_read_dict["ambiguous_cds_totals"][tran] = cds_total+ambig_cds_total master_read_dict["ambiguous_threeprime_totals"][tran] = three_total+ambig_three_total print ("Writing out to sqlite file") sqlite_db = SqliteDict(outputfile,autocommit=False) for key in master_read_dict: sqlite_db[key] = master_read_dict[key] sqlite_db["description"] = desc sqlite_db.commit() sqlite_db.close() if __name__ == "__main__": if len(sys.argv) <= 2: print ("Usage: python bam_to_sqlite.py <path_to_bam_file> <path_to_organism.sqlite> <file_description (optional)>") sys.exit() bam_filepath = sys.argv[1] annotation_sqlite_filepath = sys.argv[2] desc = sys.argv[3] outputfile = sys.argv[4] process_bam(bam_filepath,annotation_sqlite_filepath,outputfile)