Hi,
I have recently performed gene-level alignment with STAR on 20 samples with the parameter --quantMode GeneCounts and --outSAMtype BAM SortedByCoordinate. I have the output files ReadsPerGene.out.tab and Aligned.sortedByCoord.out.bam.
From this, how can I generate reliable TPM values with either the sorted BAM files or the raw counts? I saw posts indicating that TPM values can be manually calculated using the gene length from the GTF/GFF3 file. However, I am hesitant in doing this, as there might be additional factors or parameters that have been overlooked.
Currently, my last resort will be to realign using STAR again, but on transcriptomeSAM mode and then use it as input to cufflink or salmon. Just wondering if there is a more convient and reliable approach in doing this with gene-level STAR alignments. Thanks!
Ask yourself whether your analysis really needs the added precision in terms of accurate gene length from salmon/kallisto/similarTools rather than just using the length from STAR. If so then use this option to get accurate gene length estimates.
Why is that a last resort? That will generate the most reliable TPM values.