IPAFinder performs de novo identification and quantification of dynamic IpA events using standard RNA-seq, regardless of any prior poly(A) site annotation. Assuming there is an intronic poly(A) site used in a given intron, IPAFinder models the normalized single-nucleotide resolution RNA-seq read coverage profiles and identifies profound drop in coverage to infer the used poly(A) site. To detect skipped IpA, IPAFinder recognized cryptic 3′ splice site by junction-spanning reads and concatenated preceding exon to potential terminal exon. IPAFinder also has the ability to exclude alternative splicing events such as alternative 5′ splice site and cryptic exon activation by recognizing junction-spanning reads.

IPAFinder consists of both Python (3.5+) and R scripts:

1. Install the following software pre-requisites:

   i. python (required packages HTSeq, itertools, numpy, collections, multiprocessing, scipy, argparse, os, warnings, and subprocess)

   ii. R (required packages optparse, dplyr, stringr, and DEXSeq)

2. Clone the lastest development version of IPAFinder and change directory:

 git clone https://github.com/ZhaozzReal/IPAFinder.git
 cd IPAFinder

IPAFinder has three sub-commands:

1.IPAFinder_GetAnno.py: Generate annotation file containing intron and exon information

2.IPAFinder_DetectIPA.py: Detect and quantify IPA sites, and calculate read counts of all exons

3.Infer_DUIPA.R: Infer differential usage of IPA sites

RefSeq GTF file could be downloaded from the UCSC website: https://genome.ucsc.edu/.

The UCSC tool gtfToGenePred is required here.

Command

python IPAFinder_GetAnno.py -gtf /path/to/hg38refGene.gtf -output /path/to/IPAFinder_anno_hg38.txt

We have generated annotation file for hg19, hg38 and mm10, and we suggest users utilize it directly.

Command

python IPAFinder_DetectIPA.py -b /path/to/allbamfiles.txt -anno /path/to/IPAFinder_anno_hg38.txt -p 10 -o /path/to/IPAFinder_IPUI.txt

allbamfiles.txt contains all filename of bamfile between two conditions, as shown below:

condition1=/path/to/ctrl1.bam,/path/to/ctrl2.bam 
condition2=/path/to/case1.bam,/path/to/case2.bam

Following counting reads mapped to all exons, IPAFinder expects the results to be located inside its own sub-directory. For example, new generated results may appear with the following directory structure:

project/
  |-- ctrl1_exoncount.txt
  |-- ctrl2_exoncount.txt
  |-- case1_exoncount.txt
  |-- case2_exoncount.txt

DEXSeq, which is widely used for differential exon usage analysis on RNA-seq data, was applied to detect differential usage of IPA sites. This statistical framework could account for biological variability between replicates and is robust to changes in isoform abundance between conditions.

Command

Rscript Infer_DUIPA.R -b /path/to/allbamfiles.txt -I /path/to/IPAFinder_IPUI.txt -d /path/to/project -o /path/to/IPAFinder_DUIPA.txt

Final results will be saved in the file IPAFinder_DUIPA.txt.

The final output format is as follows:

Option 1: Infer differentially used IPA sites using Fisher's exact test-based method

python IPAFinder_PS_FET.py -b1 /path/to/ctrl.bam -b2 /path/to/case.bam -anno /path/to/IPAFinder_anno_hg38.txt -p 10 -o /path/to/IPAFinder_DUIPA.txt

Option 2: Infer differentially used IPA sites using bootstrapping-based method

python IPAFinder_PS_FDR.py -b1 /path/to/ctrl.bam -b2 /path/to/case.bam -anno /path/to/IPAFinder_anno_hg38.txt -p 10 -o /path/to/IPAFinder_DUIPA.txt