tincheck 0.1.0

tincheck

Tincheck

Tincheck is a python package to calculate transcript Integrity Number (TIN) and Transcription overlap.

Transcripts with transcription overlap or uneven coverage can result in false positives in differential expression analysis. Transcript Integrity Number is a metric that calculates coverage evenness and can be used as a filtering criteria in RNA-Seq studies to improve the accuracy of results.

Transcript Integrity Number (TIN) calculation

TIN denotes how evenly a feature is covered by reads. By default, the script calculate the coverage evenness across a gene by considering the coverage across all exons. However, the script can calculate coverage evenness across a transcript or any other feature that is in the annotation file.

Inputs

1. Alignment file in bam format
2. Annotation file in GTFor GFF3 format

Output

Tab delimited textfile with TIN score calculated for each gene/transcript/any other feature specified in the input. An example is given below.

target_id       eff_length  S1_count    S1_exp_tin  S1_obs_tin
PF3D7_0102700	1683	    670	        100.0	    70.9
PF3D7_0103700	1624	    135	        100.0	    72.8
PF3D7_0107300	1581	    4508        100.0	    70.4
PF3D7_0107600	5702	    4979        100.0	    74.9
PF3D7_0107800	4424	    924	        100.0	    78.8

How to install the script?

python setup.py install

It would be better to create a conda environment and install the script as shown below

conda create -n tincheck -y python=3.8
conda activate tincheck

Install additional requirements

conda install --file conda-requirements.txt 

Finally, install the script in the conda environment as

python setup.py install

How to run the script?

The required inputs to the script is bam file(s) and annotation file in GTF or GFF format.

The annotation file should have a gene feature row present in it.

Given a bam file and an annotation file, the simplest way to run the script is

tincheck tin --ann data/ann.gtf data/sample.bam 

How to calculate TIN across coding regions of a gene?

tincheck tin  --ann data/ann.gtf --feat CDS data/sample.bam

How is tin calculated

Overlapping features specified by --feat are merged together and coverage evenness is captured using Shannon's entropy formula (H). This is then convered into TIN Score as

TIN = (100*exp(H))⁄length  where H= Shannon's entropy formula.

Transcription overlaps

Transcription overlaps are flagged using overlap.py script.

tincheck overlap --ann data/ann.gtf data/sample.bam >sample_overlap.txt

By default, genes are checked for overlap only if

gene-tin < tin-cutoff and gene-count > count-cutoff

ie, genes with enough counts but low tins are checked for transcription overlap

A gene is flagged to have a transcription overlap if either of its neighboring genes and the corresponding inter-genic region have a higher tin-score and read count than the gene of interest.

In stranded mode, if the count and tin values of neighboring genes and the intergenic region in the gene sense strand is comparable to the gene count and tin, then that gene is considered to have an overlap from neighboring transcripts.

More specifically in stranded mode, a gene is considered to have a neighboring transcript overlap, if either of the following conditions are satisfied.

1. left gene tin in gene sense strand >= gene tin and left intergenic tin in gene sense strand >=gene-tin
2. right gene tin in gene sense strand >= gene tin and right intergenic tin in gene sense strand >=gene-tin