LIQA2

LIQA2 is a computational tool for detecting and quantifying isoform expression (including novel isoforms) from long read sequencing. The tool uses a meta-regression model to quantify differential alternative splicing and incorporates a targeted approach to correct for 3' and 5' biases.

Installation

git clone https://github.com/WGLab/LIQA2.git
cd LIQA2

Pipeline

The main function for running the LIQA2 pipeline is main_preprocessing.py. There are 4 steps, marked by the --task [step] command.

1. Prepare annotation file

The first step of the LIQA2 pipeline involves generaing annotation files from the reference genome and indexed/tagged BAM files. This is done with --task annotation.

Arguments include:

• --target: path to output location(s)
• --bam: path to bam location(s)
• annotation mode (see below)
• --reference: path to GTF annotation file
• reference_pkl: path to picke gene annotation file generated by LIQA2 (geneStructureInformation.pkl)
• --workers: threads for parallel processing
• --coverage_threshod_exon: coverage threshold for exon discovery as a percent of maximum coverage. Default is 0.02, larger values are more conservative

Below is a sample command that can be run.

python3 src/main_preprocessing.py \
--task annotation \
--target path/to/output/folder/of/sample1 path/to/output/folder/of/sample2 \
--bam path/to/bam/file/or/bamfolder/sample1 path/to/bam/file/or/bamfolder/sample2 \
--update_gtf \
--reference path/to/reference/genes.gtf \
--reference_pkl data/geneStructureInformation.pkl \
--workers 30

There are 3 different modes of annotation.

1. Annotation-Only:
2. Enhanced-Annotation:
3. Annotation-Free:

2. Generate compatible matrix

This step creates a compatible matrix that aligns reads to existing gene isoforms and identifies novel isoforms based on previously generated annotation files. Arguments include:

• --target: path to output location(s)
• --bam: path to bam location(s)
• --reference: path to GTF annotation file

Below is a sample command that can be run.

python3 src/main_preprocessing.py \
--task 'compatible matrix' \
--target path/to/output/folder/of/sample1 path/to/output/folder/of/sample2 \
--bam path/to/bam/file/or/bamfolder/sample1 path/to/bam/file/or/bamfolder/sample2 \
--reference path/to/reference/genes.gtf

3. Summarize novel gene annotations

After annotation and compatible matrix generation, the LIQA2 pipeline generates a new summary annotation file that also includes novel isoforms:

python3 src/main_preprocessing.py \
--task 'summary' \
--target path/to/output/folder/

4. Generate count matrix

The final step generates a count matrix. Arguments include:

• --target: path to output location(s)
• --bam: path to bam location(s)
• --reference: path to GTF annotation file
• --workers: threads for parallel processing -- save_csv/--save_mtx: set up output format --group_novel: whether group some novel isoforms that are potentially generated by read truncations together as one novel isoform. --platform: 10x-ont or 10x-pacbio

Below is an example:

python3 src/main_preprocessing.py \
--task 'count matrix' \
--target path/to/output/folder/ \
--platform 10x-ont
--workers 8 --group_novel

With these steps, you should be able to generate counts that you can use for expression analysis or further study.

Citation

Gouru, A., Xu, Z., Wang, K. “LIQA2: Efficient Isoform Detection and Quantification from Long Read Sequencing,” October 2024. International Conference on Intelligent Biology and Medicine, Houston, TX.