Counts in fractions.tsv

[linkerst]

Dear Yuanhua,

This is more a question than an issue. I recognized that the counts in the fractions.tsv file can adopt non integer values, because the counts are the mean of Cnt_all .

Cnt_all is updated in your MH_propose function together with the Y and Psi value. Could you please describe shortly the connection between the actual counts in the sequencing data and your Cnt_all? Unfortunately I am not able to figure it out by looking at your code our paper.

Many thanks, Stephanie

[huangyh09]

Dear Stephanie,

Thanks for the good question, and checking the codes. As inputs, the counts for a whole splicing event are integer values, however, when assigning these reads to either exon-inclusion isoform or exon-exclusion isoform, it takes the posterior probability, which is a float. That is why you see them as non integer values. When you add up the counts for the two isoform, i.e., xx.in and xx.out, it will be very close to an integer. For simplicity, you could take the float value as an integer with round() function, which may sacrifice some precision, and is taken for the output file of differential splicing by brie-diff.

In the script, the Cnt_all is the counts supporting each isoform for all splicing events. This is based on the estimate of the fractions of each isoform (within a gene), i.e. Psi, and then multiply to the total counts of that gene, with weights of isoform length.

So, the counts are very informative, and is good for using as threshold when selecting genes for analysis, though the confidence interval is also a good option.

I will add more details on the output counts into the manual.

Cheers, Yuanhua

[aarzalluz]

Hi Yuanhua,

Has the manual been updated yet with more information about how the count values mentioned in the first section of the manual are calculated? I'm interested in understanding the tool better, as I'm considering using it for my own analyses, but quantification methods are crucial, and I cannot find details about them anywhere in the documentation.

Cheers,

Ángeles