Gene location-aware accurate quantitation of gene expression from high-throuput sequencing alignments.

pip install -e git+https://github.com/mikpom/uslcount.git#egg=uslcount

--prefix option of pip install can be used to specify custom installation root directory.

Add --upgrade to upgrade existing installation.

Package can be used to quantitate reads by gene features from BAM alignments. For unstranded libraries on top of gene expression characteristics package can give the confidence score idicating how likely the counts are compromised (overestimated) due to expression from the opposite strand. One might need to manually inspect those with low confidence score and maybe consider using splice junction counts (also contained in output file) as an expression estimate of these genes.

It is assumed that NGS reads were already aligned to the genome using aligner of choice (e.g. STAR). Prior to running quantitation analysis a genomic database should be created which will be then used for processing of BAM files.

Available tasks can be listed by typing:

python3 -m uslcount

To build genomic database run

python3 -m uslcount build /path/to/GTF /path/to/GenomicDataDir

To count reads run

python3 -m uslcount count -out outfile GenomicDataDir BAM

where GenomicDataDir indicated path to data directory created at the build step.

To see other options

python3 -m uslcount count -h

To convert count values to TPM one needs to know gene length parameter to make comparisons between different genes more adequate. Actual span from which counts are gathered is dependent on whether the analysis was stranded or not. If counting was unstranded then corresponding gene lengths can be obtained from the file len_ust_unq. For stranded analysis lengths are stored in the file len_st_unq. The latter lengths are equal or larger than the former. Both files can be found in genomic directory indicated at the build step.

Note that for some genes length equals to zero meaning there is no non-overlapping exons to distinguish that gene from the other. In this case count is always zero.

To get confidence score of counts obtained from unstranded libraries analyze task should be invoked. Usage is very similar to count task:

python3 -m uslcount analyze -out outfile GenomicDataDir BAM

This will output counts, junction counts and a confidence score (int from 1 to 10) indicating how robust expression value is. 10 means expression value is reliable, 1 means -- least reliable. Junction counts can be used as a proxy for gene count in differential expression analysis in those cases when gene's expression biased by opposite strand transcription.

Please cite the following if you find the package useful.

Pomaznoy, M., Sethi, A., Greenbaum, J. et al. Identifying inaccuracies in gene expression estimates from unstranded RNA-seq data. Sci Rep 9, 16342 (2019). https://doi.org/10.1038/s41598-019-52584-w