SAJR

SAJR is R-package to perform alternative splicing quantification and differentiall splicing analysis using bulk RNA-seq data. SAJR pipeline consists of two steps:

1. Preparation of annotation. SAJR uses annotation in its own gff-based format. To convert gtf or gff3 into format acceptable use gff2sajr or gff32sajr methods (see below).
2. Counting reads using java-written read-counter. It takes bams and annotation in gff format as input and produces text files with read counts.
3. Differentiall (and other downstream) analysis using R-package from this repository. It take gtf and output from read-counter as input.

Installation

R-package can be installed from within R-session by

devtools::install_github('iaaka/sajr')

The precompiled read-counter jar and config can be downloaded from github by

wget https://github.com/iaaka/sajr-java/raw/main/sajr.jar
wget https://github.com/iaaka/sajr-java/raw/main/sajr.config
java -jar sajr.jar

Last command should print short help message, if it works then sajr was downloaded succesfully.

Testing

Download test data:

git clone https://github.com/iaaka/sajr-java.git
mv sajr-java/example .
rm -rf sajr-java

Then run sajr read counter on toy dataset:

java -jar sajr.jar count_reads

It will use parameters, input, and output files specified in sajr.config that points to data in example folder. It should produce .gene, .seg, and .intron files for each of two input files (six in total).

How it works

SAJR works on level of segments - parts of gene between two closest splicing sites or TSS or polyA-sites. Then for each segment it counts number of reads that supports its inlusion (reads that overlap the segment, inclusion reads) and these that supports its exclusion (reads mapped to exon-exon junctions that span the segment, exclusion reads).

SAJR normalize these counts by read and segment length to calculate psi (percent spliced in):

SAJR uses generalized linear models (glm) with quasibinomial distribution (to acccount for biological variability) for differential splicing test.

Annotation conversion

To split intput gene annotation in gtf or gff3 format into segments one of command below can be used, use gtf option if possible as more stable:

java -jar sajr.jar gff2sajr -ann_foreign=input.gtf -ann_out=output.sajr

or

java -jar sajr.jar gff32sajr -ann_foreign=input.gff3 -ann_out=output.sajr

Read-counter

Read-counter takes converted annotation and bam files as input and returns. The parameters can be provided either in config file or redefined in command line. By default config file should be named sajr.config and placed in the same folder as sajr.jar, or it can be specified by command line parameters:

java -jar sajr.jar count_reads my.config \
  -batch_in=example/1.bam \
  -batch_out=out1 \
  -ann_in=example/a.gff \
  -count_only_border_reads=true

For parameters please see detailed manual.

Sajrcomp

Sajrcomp takes two annotation files in SAJR-format (gff), provides comparing of features and returns the counts of features overlapping.

java -jar sajr.jar sajrcomp \
  -comp_ann1=ann1 \
  -comp_ann2=ann2 \
  -comp_out=output.tsv

Output A signle file with four columns:

gene_id1 - gene identifier from comp_ann1
gene_id2 - gene identifier from comp_ann2
class - class of overlap (see below)
number of introns: in gene from comp_ann1, from comp_ann2 and common 

Overlap classes

= - both genes have exactly the same set of junctions
c - junctions of one gene are subset of junctions of other one
j - genes share some junctions
e - genes overlap by exons in sense
i - one gene is within intron of other one (in sense)
a - one gene is within intron of other one (in antisense)
x - genes overlap by exons in antisense

R-package

See tutorial.

Citing SAJR

• P. Mazin, J. Xiong, X. Liu, Z. Yan, X. Zhang, M. Li, L. He, M. Somel, Y. Yuan, Y.P. Chen, N. Li, Y. Hu, N. Fu, Z. Ning, R. Zeng, H. Yang, W. Chen, M. Gelfand and P. Khaitovich. Widespread splicing changes in human brain development and aging. Molecular Systems Biology 9:633. doi