can someone please explain the DAPARS output?

Question

[PK]

sorry for the primitive question, I'm quantifying the poly usage from the poly-A selected RNA SEQ library. i'm using DAPARS to do the job and i got the output but i got stuck here. The output give

A_1_long_exp A_1_short_exp A_1_PDUI B_1_long_exp B_1_short_exp B_1_PDUI Group_A_Mean_PDUI Group_B_Mean_PDUI PDUI_Group_diff P_val

I have few questions here. 1) can i compare directly the short_exp of condition A and B. 2) How do i know whether the short/long UTRs used in let's say in condition B

Thanks in advance

Answer 1

[michael.ante]

Hi PK,

I'd stick to the adjusted P_val field, since the tool tests for a shift in PAS-usage. Of course you can compare the short_exp directly, but the change might or might not relative to the overall expression difference due to the different conditions.

The PDUI is more or less defined as long_exp / (long_exp + short_exp) . All PDUIs below 0.5 use more the shortened transcript.

I hope this helps.

Cheers,

Michael

[PK]

Great. Thanks for your reply. Also do i have to rely on the filter=yes. Because i have triplicates(WT vs KO). After the analysis non of the transcript passed the filter. I have ~200 million reads per library.

[michael.ante]

The filter, according to their paper checks if the FDR is <= 0.05 AND the delta PDUI is >= 0.2 AND the log2 fold-change in PDUI is >= 0.59 . You might want to check why your data lacks these criteria. For instance select all items with FDR <= 0.05 and delta-PDUI >=0.2 and check how the log2 fold-change behaves.

Alternatively, you can check a different tool like apa

[PK]

Thanks for the comment. yeah it looks not good. For some reason the delta-PDUI >=0.2 is not so great (i have lot of negative values). correct me if i'm wrong, in the paper it is mentioned del-PDUI can range from -1 to +1. Does negative del-PDUI corresponds to the shorter UTR? I will use APA and thanks for you suggestion. Also, i have this doubt with me. While crating the wig files i used effective genome size as a normalization factor and the bin size is 10 . do you think any of these parameters could affect the output.