Different length and counts depending on the number of loaded samples

[FlorianRocher]

Hi,

I noticed something weird using tximport by summarizing at the gene level.
Depending on the number of samples I load I don't have the same length and counts for my genes in the corresponding samples.

Quant_Matrix_tximport=tximport(**AlltheFiles**,txIn = T, txOut = F, type="salmon", countsFromAbundance ="lengthScaledTPM", tx2gene = tx2gene)
Quant_Matrix_tximport1=tximport(**20FirstFiles**,txIn = T, txOut = F, type="salmon", countsFromAbundance ="lengthScaledTPM", tx2gene = tx2gene)

Example:
When all samples are load I have Sample1:Gene 1 --> Counts=386.3943113
When 20 samples are loaded I have Sample1:Gene1 --> Counts=383.4657441

I used "no" option insted to check. I did not see any differences at the count level but at the length of the genes level.

Is this something that is expected ?

Florian Rocher

[mikelove]

Yes expected.

When a gene has some samples with no expression for any isoform, the average length for those samples depends on the other samples.