barcodex 1.0.3

A tool for extracting Unique Molecular Identifiers (UMIs) from single or paired-end read sequences

BarcodEX

BarcodEX is tool for extracting Unique Molecular Identifiers (UMIs) from single or paired-end read sequences. It can handle UMIs inline with reads or located in separate fastqs.

Installation

From PyPi

BarcodEx is available from PyPi

pip install barcodex

From GitHub

Clone the BarcodEx repository

git clone https://github.com/oicr-gsi/barcodex

Install required python libraries by running pip install -r requirements.txt

Extraction of UMI sequences in extract mode

Parameters

argument	purpose	required/optional
--r1_in	Path(s) to FASTQ(s) containing read 1	required
--r1_out	Path to output FASTQ 1	required
--r2_in	Path(s) to FASTQ(s) containing read 2	optional
--r2_out	Path to output FASTQ 2	optional
--r3_in	Path(s) to FASTQ(s) containing UMIs for paired end with non-inline UMIs	optional
--pattern1	pattern or regex for extracting UMIs in read 1	optional
--pattern2	pattern or regex for extracting UMIs in read 2	optional
--data	paired or single end sequencing	required
--separator	String separating the UMI sequence in the read name	required
--umilist	Path to file with valid UMIs	optional
--inline	UMIs inline with reads or not	optional
--keep_extracted	Writes discarded and UMIs sequences to file	optional
--keep_discarded	Writes reads without matching pattern to file	optional
--full_match	Requires the regex pattern to match the entire read sequence	optional
--compressed	Compresses output fastqs with gzip	optional

BarcodEx extracts UMIs using either a pattern sequence or a regular expression and appends the concatenated UMIs to the read name preceded by a separator string specified in the command. UMIs can be extracted from read 1 and/or read 2 using respectively --pattern1 and --pattern2. At leat 1 pattern must be used. When extracting UMIs in read 1 and read 2, --pattern1 and --pattern2 must be both either a string sequence or a regular expression. Reads that are not matching the provided patterns are discarded. Discarded reads can be recovered for inspection. Morover, the extracted sequences can also be recovered and written to file if the original fastqs need to be re-generated (see below).

Extraction with a string pattern

The pattern sequence must include one or more Ns, indicating the UMI bases, optionally followed by any nuccleotides corresponding to spacer sequence. For instance the pattern NNNNN extracts the first 5 nucleotides from the read whereas pattern NNNNNATCG extracts the first 9 nucleotides, appends nucleotides 1-5 to the read name and discard spacer ATCG. Reads not matching NNNNNATCG are discarded.

Extraction with the pattern sequence always extracts UMIs at the beginning of the read sequence. Extraction with regular expression offers more flexibility in the UMI design (see below).

As an example, consider read:

@MISEQ753:39:000000000-BDH2V:1:1101:17521:1593 1:N:0:
TCATGTCTGCTAATGGGAAAGAGTGTCCTAACTGTCCCAGATCGTTTTTTCTCACGTCTTTTCTCCTTTCACTTCTCTTTTTCTTTTTCTTTCTTCTTCTT
+
1>1A1DDF11DBDGFFA111111D1FEEG31AD1DAA1110BA00000//01A2A/B/B/212D2111D1222D12122B1B01D1@101112@D2D12BB

Extraction with pattern NNNNNNNNNNNNATGGGAAAGAGTGTCC will extract UMI TCATGTCTGCTA and add it to the read name. Spacer sequence ATGGGAAAGAGTGTCC is removed from read. So the new read is now:

@MISEQ753:39:000000000-BDH2V:1:1101:17521:1593_TCATGTCTGCTA 1:N:0:
TAACTGTCCCAGATCGTTTTTTCTCACGTCTTTTCTCCTTTCACTTCTCTTTTTCTTTTTCTTTCTTCTTCTT
+
G31AD1DAA1110BA00000//01A2A/B/B/212D2111D1222D12122B1B01D1@101112@D2D12BB

Extracted sequence TCATGTCTGCTAATGGGAAAGAGTGTCC and its corresponding qualities 1>1A1DDF11DBDGFFA111111D1FEE can be written to fastq file with option --keep_extracted (see below).

Extraction with a regular expression

Regular expressions allow more flexibility for extracting UMIs, in particular UMIs with complex design and UMIs not starting at the beginning of the read. A good introduction to regular expression can be found in this Regular Expression HOWTO. BarcodEx depends on the regex module rather than the standard re module because the former allows fuzzy matching.

Sequences are extracted from the read using named groups within the regex. Allowed named groups are umi and discard. Syntax with named groups is as follow: (?<umi>.{3})(?<discard>T{2}): extracts a 3bp UMI followed by TT spacer that is removed from read and discarded The discard group removes nucleotides and qualities from the read while the umi group extracts the UMI that gets added to the read name. Any sequence not contained in umi and discard groups will remain in the read. Thus, it is important to construct the regular expression such that the begining of the read is captured in groups.

For instance, consider the following read:

@MISEQ753:39:000000000-BDH2V:1:1101:17521:1593 1:N:0:
AATCGTCCATCG
+
1>1A1DDF11DB

The regex (?<umi>.{3})(?<discard>C{2}) will extract UMI CGT and discard spacer CC. But the first 3 nucleotides AAT will remain in the read with new read being:

@MISEQ753:39:000000000-BDH2V:1:1101:17521:1593_CGT 1:N:0:
AATATCG
+
1>111DB

To prevent the AAT before the UMI CGT to be part of the read, we need to account for the nucleotides upstream in the UMI in the regex (?<discard1>^.*)(?<umi>.{3})(?<discard2>C{2})

@MISEQ753:39:000000000-BDH2V:1:1101:17521:1593_CGT 1:N:0:
ATCG
+
11DB

The extracted sequences and qualities can be recovered with option --keep_extract which writes the following read to file (see below).

@MISEQ753:39:000000000-BDH2V:1:1101:17521:1593_CGT 1:N:0:
AATCGTCC
+
1>1A1DDF

Multiple umi and discard named groups are allowed within the regex but they should be named differently. Naming is not important as long as groups contain the strings umi and discardwithout special characters. For instance the following 2 regex will give the same output:

• (?<discard_1>^.*)(?<umi>.{3})(?<discard_a>C{2})') and (?<discard1>^.*)(?<umi>.{3})(?<discard2>C{2})')
• (?P<umi_1>^[ACGT]{3}[ACG])(?P<discard_1>T)|(?P<umi_2>^[ACGT]{3})(?P<discard_2>T) and (?P<umi_a>^[ACGT]{3}[ACG])(?P<discard_a>T)|(?P<umi_b>^[ACGT]{3})(?P<discard_b>T)

Filtering extracted UMIs against a list

Extracted UMIs can be filtered out against a list of validated UMIs provided as a table file with --umilist. The UMIs must be in the first column and any other columns are ignored. Reads for which the extracted UMIs are not in the list are discarded. For paired end reads, both reads are discarded if any UMI is not in the list when UMIs are extracted from each read.

Extraction of UMIs in single or paired read sequences

Single and paired end read data are indicated with the option --data single or --data paired respectively. For paired end data with inline UMIs, options --r1_in and r2_in indicate the paths to the read 1 and read 2 fastqs and --r1_out and r2_out indicate the paths to the read1 and read2 output fastqs. Only r1_in and r1_out are required for single end data with inline UMIs. Output fastqs are compressed if --compressed is used. The .gz extension is added to --r1_out and/or --r2_out if not specified. Input fastqs can be compressed with gzip or uncompressed. Input fastqs for paired end data must be in sync.

Extraction from multiple input fastqs

Multiple input fastqs can be processed together for read 1 and/or read 2 but generating a single output fastq for single end data and 2 output fastqs for paired read data. The files mut be passed to --r1_in for read 1 fastqs and --r2_in for read 2 fastqs, each file being separated by white space. The number of input fastqs for paired data must be the same for read 1 and read 2 and each list of files must be in the same order.

Extraction of UMIs not inline with reads

With option --inline, BarcodEx expects UMIs to be inline with the read. For some library types, such as sureselect and haloplex, the UMIs are not inline but are located in a separate fastq. Omitting --inline assumes UMIs to be in a fastq file. This file is indicated with --r2_in for single end data and --r3_in for paired end data. With UMIs in file, --pattern1 is used to extract UMIs from --r2_in or --r3_in and --pattern2 is not used.

Recovering discarded reads and extracted sequences

Reads without a matching pattern can be written to file for inspection with option --keep_discarded. The fastqs with discarded reads are written in the same directory as r1_in. File names with discarded reads are modeled after --r1_out and --r2_out with suffix ".discarded.R1/2.fastq". Extracted read sequences (UMIs and any spacer sequence removed from read, along with their qualities) can also be written as a fastq file with option --keep_extracted. This allows to re-generate the original fastqs using the r1_out and/or r2_out fastqs together with the fastqs with extracted reads. The fastqs with extracted reads are written in the same directory as r1_in. File names with extracted reads are modeled after --r1_out and --r2_out with suffix ".extracted.R1/2.fastq" for inline UMIs and ".extracted.R2/3.fastq" for UMIs located in files.

UMIs and reads metrics

Two files with metrics of the UMI extraction are written in json format in the same directory as --r1_out. The files are named after --r1_out with suffix "_extraction_metrics.json' and "_UMI_counts.json". The first json captures information about the extraction process:

• total reads/pairs processed
• number reads/pairs with matching pattern
• numberreads/pairs with non-matching pattern
• pattern1
• pattern2
• umi-list

The second json records the UMI counts after extraction. For paired end data it counts the concatenated sequences with UMIs from read 1 and read 2, but adding a "." separator to track the read origin of each UMI. For instance, "AAA.TCG": 10 in the json file indicates that sequence AAATCG is found 10 times in all the extracted UMIs, and that it is made of AAA from read 1 and TCG from read 2. One can then easily obtain counts of all UMIs from read 1 and read 2.

Importing Barcodex as a module

BarcodEx can be run as a script or imported as a module to perform extraction within your own script. The recommended import is from barcodex import extract_barcodes. Dependent modules must also be imported (see Example command #8 below).

from barcodex import extract_barcodes

help(extract_barcodes)
Help on function extract_barcodes in module barcodex:

extract_barcodes(r1_in, r1_out, pattern1, pattern2=None, inline_umi=True, data='single', keep_extracted=True, keep_discarded=True, r2_in=None, r2_out=None, r3_in=None, full_match=False, separator='_', compressed=True, umilist=None)
    (list, str, str | None, str | None, bool, str, bool, bool, list | None, str | None, list | None, bool, str, bool, str | None) -> None

    Parameters
    ----------

    - r1_in (list): Path(s) to the input FASTQ 1 (compressed or not)
    - r1_out (str): Path to the output FASTQ 1 with reads re-headered with UMI sequence 
    - pattern1 (str or None): String sequence or regular expression used for matching and extracting UMis from reads in FASTQ 1.
                             The string sequence must look like NNNATCG or NNN. UMI nucleotides are labeled with "N".
                             Spacer nucleotides following Ns are used for matching UMIs but are discarded from reads    
                             None if UMIs are extracted only from FASTQ 2 for paired end sequences
    - pattern2 (str or None): String sequence or regular expression used for matching and extracting UMis from reads in FASTQ 2.
                             The string sequence must look like NNNATCG or NNN. UMI nucleotides are labeled with "N".
                             Spacer nucleotides following Ns are used for matching UMIs but are discarded from reads    
                             None if UMIs are extracted only from FASTQ 1 for paired end sequences
    - inline_umi (bool): True if UMIs are inline with reads and False otherwise
    - data (str): Indicates if single or paired end sequencing data
    - keep_extracted (bool): Write extracted sequences (UMIs and discarded sequences) to file if True
    - keep_discarded (bool): Write reads without matching pattern to file if True
    - r2_in (list or None): Path(s) to the input FASTQ 2 (compressed or not)
    - r2_out (str or None): Path to the output FASTQ 2 with reads re-headered with UMI sequence    
                           None for single end read sequences
    - r3_in (list or None): Path(s) to input FASTQ 3 for paired end sequences with non-inline UMIs 
    - full_match (bool): True if the regular expression needs to match the entire read sequence
    - separator (str): String separating the UMI sequence and part of the read header
    - compressed (bool): output fastqs are compressed with gzip if True
    - umilist (str or None): Path to file with accepted barcodes. Barcodes are expected
                               in the 1st column. Any other columns are ignored.

Example commands

Example 1. Paired end end with string sequence

Extraction of UMIs in read 1 for paired end data with inline UMIs with a string sequence. Extracts the first 12 bp UMI when followed by spacer ATGGGAAAGAGTGTCC and remove spacer from read. All output fastqs are compressed. Discarded reads and extracted sequences are recovered. Umis are preceded by an underscore in the read header.

python3.6 barcodex.py extract \
--r1_in myfile_R1.fastq \
--r1_out output_R1.umis.fastq.gz \
--r2_in myfile_R2.fastq \
--r2_out outout_R2.umis.fastq.gz \
--inline --data paired \
--pattern NNNNNNNNNNNNATGGGAAAGAGTGTCC \
--separator "_" --keep_extracted \
--keep_discarded --compressed

Example 2. Paired end with regex

Extraction of UMIs in read 1 and read 2 for paired end data with inline UMIs with a regular expression. Extracts the first 3 nucleotides as UMI and discards the next 2 nucleotide spacer sequence from each read. All output fastqs are compressed. Discarded reads and extracted sequences are recovered. Umis are preceded by an underscore in the read header.

python3.6 barcodex.py extract \
--r1_in myfile_R1.fastq.gz \
--r1_out output_R1.umis.fastq.gz /
--r2_in myfile_R2.fastq.gz \
--r2_out outout_R2.umis.fastq.gz \
--inline --data paired --compressed \
--pattern1 "(?<umi>.{3})(?<discard>.{2})" \
--pattern2 "(?<umi>.{3})(?<discard>.{2})" \
--separator "_" --keep_extracted --keep_discarded 

Example 3. Full match regex option

Same as Example 2, but the regex patterns are modified to suit the --full_match regex requirement.

python3.6 barcodex.py extract \
--r1_in myfile_R1.fastq.gz \
--r1_out output_R1.umis.fastq.gz \
--r2_in myfile_R2.fastq.gz \
--r2_out outout_R2.umis.fastq.gz \
--inline --data paired --compressed \
--pattern1 "(?<umi>.{3})(?<discard>.{2}.+)" \
--pattern2 "(?<umi>.{3})(?<discard>.{2}.+)" \
--separator "_" --keep_extracted \
--keep_discarded  --full_match

Example 4. List of UMIs

Extraction of UMIs in read 1 and read 2 for paired end data with inline UMIs with a regular expression. Extracts a 4 bp UMI not ending with T and discard a following T spacer or a 3 bp UMI and discard the following T spacer. UMIs start at the beginning of the read sequence and only higher caps A, T, C and G are allowed. Extracted UMIs are checked against the true_barcode.txt UMI list. Reads with non-valid UMIs (ie. not present in true_barcode.txt) are discarded. All output fastqs are compressed. Discarded reads and extracted sequences are recovered. Umis are preceded by an underscore in the read header.

python3.6 barcodex.py extract \
--r1_in myfile_R1.fastq.gz \
--r1_out output_R1.umis.fastq.gz \
--r2_in myfile_R2.fastq.gz \
--r2_out outout_R2.umis.fastq.gz \
--inline --data paired --compressed
--separator "_" --keep_extracted --keep_discarded \
--umilist true_barcodes.txt
--pattern1 "(?P<umi_1>^[ACGT]{3}[ACG])(?P<discard_1>T)|(?P<umi_2>^[ACGT]{3})(?P<discard_2>T)" \
--pattern2 "(?P<umi_1>^[ACGT]{3}[ACG])(?P<discard_1>T)|(?P<umi_2>^[ACGT]{3})(?P<discard_2>T)"

example 5. Single end

Extraction of UMIs in read 1 for single end with inline UMIs with a regular expression. Extracts the first 12 bp UMI when followed by spacer ATGGGAAAGAGTGTCC and remove spacer from read. Output fastq is uncompressed. Discarded reads and extracted sequences are not recovered. Umis are preceded by an underscore in the read header.

python3.6 barcodex.py extract \
--r1_in myfile_R1.fastq.gz \
--r1_out output_R1.umis.fastq \
--inline --data single \
--pattern1 "(?P<umi>^.{12})(?P<discard>ATGGGAAAGAGTGTCC)" \
--separator "_"

Example 6. Multiple input fastqs

Extraction of UMIs in read 1 and read 2 for paired end data with inline UMIs with a regular expression from 4 input fastq1 and 4 input fastq2. Extracts the first 3 nucleotides as UMI and discards the next 2 nucleotide spacer sequences from each read. All output fastqs are compressed. Discarded reads and extracted sequences are not tracked.
Umis are preceded by an underscore in the read header.

python3.6 barcodex.py extract \
--r1_in myfile_R1.1.fastq.gz myfile_R1.2.fastq.gz myfile_R1.3.fastq.gz myfile_R1.4.fastq.gz 
--r1_out output_R1.umis.fastq.gz 
--r2_in myfile_R2.1.fastq.gz myfile_R2.2.fastq.gz myfile_R2.3.fastq.gz myfile_R2.4.fastq.gz 
--r2_out output_R2.umis.fastq.gz 
--inline --data paired \
--pattern1 "(?<umi>.{3})(?<discard>.{2})" \
--pattern2 "(?<umi>.{3})(?<discard>.{2})" \
--separator "_" --compressed

Example 7. Paired end with UMIs in file

Extraction of UMIs in read 1 and read 2 for paired end data with UMIs located in read 3 with a regular expression. Extracts the first 10 nucleotides as UMI. All output fastqs are compressed. Discarded reads and extracted sequences are recovered. Umis are preceded by an underscore in the read header.

python3.6 barcodex.py extract \
--r1_in myfile_R1.fastq.gz \
--r1_out output_R1.umis.fastq.gz \
--r2_in myfile_R2.fastq.gz \ 
--r2_out output_R2.umis.fastq.gz \
--r3_in myfile_R3.fastq.gz \ ## file with UMIs
--separator "_" --keep_extracted \
--keep_discarded --compressed
--pattern1 "(?<umi>^.{10})" \
--data paired 

Example 8. Importing BarcodEx as module within a script

Sample as Example 2.

import gzip
import os
from itertools import zip_longest
import regex
import json
import time
from barcodex import extract_barcodes


extract_barcodes(['myfile_R1.fastq.gz'], 'output_R1.umis.fastq.gz', pattern1='(?<umi>.{3})(?<discard>.{2})',
                   pattern2='(?<umi>.{3})(?<discard>.{2})', inline_umi=True, data='paired',
                   keep_extracted=True, keep_discarded=True, r2_in=['myfile_R2.fastq.gz'],
                   r2_out='output_R2.umis.fastq.gz', r3_in=None, full_match=False,
                   compressed=True, umilist=None)