Cash in on Expressed Barcode Tags (EBTs) from NGS Sequencing Data

NOTE: For an updated, optimized version of cashier that is still being currently maintained, use pycashier

Cashier is a tool to simplify extraction and error-correction of expressed barcode tags from sequencing reads.

It is basically a wrapper for cutadapt an impressive program that identifies and removes flanking adapter sequences to extract the barcodes, and starcode, a radically fast minimum levenshtein clustering tool, to compensate for sequencing error in the extracted barcodes (and UMIs).

We use it to process our barcode sequencing data from both amplicon and scRNAseq sources, and it can even work with 5' UMI adapters.

This software defaults to our particular brand of crispr/cpf1 library, but it's easily relatable to your line of work.

Dependencies Needed In Path

• cutadapt - to identify and mask adapter sequences
• fastq_quality_filter - to filter minimum base quality
• starcode - to cluster UMIs and barcodes using a levenshtein distance network and message passing clustering (min centroid-point clustering threshold at default 5 to 1)

Getting Started With Amplicon Data

To extract barcode sequences from amplicon data (our crispr gRNA barcode adapters as default):

cashier_extract -i 1K.raw.fastq 

Adapted for your adapters:

cashier_extract -i 1K.raw.fastq -u <upstream adapter sequence> -d <downstream adapter sequence> 

To extract barcode and UMI sequences from amplicon data:

cashier_extract -i 1K.raw.fastq --umi 

Error Correction of UMI and Barcode Sequences

To cluster UMI and barcode sequences for a particular sample:

cluster_umi_barcode_file.sh -i sample_1.umi.barcode.tsv --distance 1 

Right now we build quite a few intermediate files, so please bear with us.

<sample_name.umi.barcode.tsv> has your groceries.

casher_extract uses cutadapt to identify and trim flanking adapters:

cashier_extract -i <input_file> 

-V show version 

-v verbose mode  (shows all bash executions) 

-i | --input  <input fastq file> 

-u | --upstream_sequence <DNA sequence upstream or 5' of barcode on sequencing read> 

-d | --downstream_sequence <DNA sequence downstream or 3' of barcode on sequencing read> 

-bl | --barcode_length <integer length of expected barcode (20)> 

-e | --error_rate allowed error tolerance in adapters  (0.1)

-q | --min_quality  <minimum PHRED quality required of all barcode bases> 
  
-umi | --umi  perform UMI extraction as well for UMIs 5' of upstream sequence 

-ul | --umi_length  <integer length of expected UMI (16)> 

-t | --trim  <integer # bases to trim on 5' end of read until UMI sequence (0)> 
  
-ul | --unlinked_adapters perform adaptor sequence trimming sequentially

-j | --threads number of cpu cores to use (1)

DEFAULTS to the following properties sutiable for our patented tool "COLBERT", a transduced expressed gRNA barcode library developed by Aziz Al'Khafaji in the Brock Lab to trace cell lineages:

upstream (5') adapter sequence: ATCTTGTGGAAAGGACGAAACACCG

downstream (3') adapter sequence: GTTTTAGAGCTAGAAATAGCAAGTT

barcode length: 20bp

umi length: 16bp

cluster_columns.sh uses starcode to run levenshtein-distance clustering on a file, then join those cluster centroids with the original data

cluster_columns.sh -i <input readname-umi-barcode.tsv file> 

-i | --input <tab-separated input input file> 

-c | --columns  the columns you want to cluster (can be comma-saparated list) 

-V | --version   show version and exit 

-v | --verbose   verbose mode - all bash executions will be printed

-d | --distance  <int minimum levenshtein distance for clustering> 

-uo | --umi-only skip barcodes when performing clustering

-t | --threads  number of cpu cores to use (1)  

-s | --spherical  use spherical rather than message-passage clustering