Number of modifications halfed in JACUSA 2.0 vs. 1.3 (mode call-2)

[ckuenne]

I noticed that the number of modifications was more than twice as high in 1.3 when compared to 2.0. Both were run on the same BAM files with equivalent parameters.

1.3.0:
call-2 -a D,M,Y --filterNM_1 5 --filterNM_2 5 -s -c 2 -T 1 -P FR-SECONDSTRAND,FR-SECONDSTRAND -p 16 -r hypertribe/jacusa/jacusa_ht_vs_m3d-ht.org ./star_ht_1/igv/ht_1.bam,./star_ht_2/igv/ht_2.bam,./star_ht_3/igv/ht_3.bam,./star_ht_4/igv/ht_4.bam ./star_m3d-ht_1/igv/m3d-ht_1.bam,./star_m3d-ht_2/igv/m3d-ht_2.bam,./star_m3d-ht_3/igv/m3d-ht_3.bam,./star_m3d-ht_4/igv/m3d-ht_4.bam
mods: 537341

2.0.1:
call-2 -a D,M,Y -filterNM 5 -s -c 2 -T 1 -P FR-SECONDSTRAND -p 16 -r hypertribe/jacusa/jacusa_ht_vs_m3d-ht.org ./star_ht_1/igv/ht_1.bam,./star_ht_2/igv/ht_2.bam,./star_ht_3/igv/ht_3.bam,./star_ht_4/igv/ht_4.bam ./star_m3d-ht_1/igv/m3d-ht_1.bam,./star_m3d-ht_2/igv/m3d-ht_2.bam,./star_m3d-ht_3/igv/m3d-ht_3.bam,./star_m3d-ht_4/igv/m3d-ht_4.bam
mods: 201810

This is before any downstream filtering. The changelog found here (https://github.com/dieterich-lab/JACUSA2) does not seem to explain that change. Most of the "lost" modifications have low ratios of modification (e.g. A 22, G 1). Is there a new filter that is not documented yet? Or did I just miss it?

[piechottam]

Do you mind sharing your "filtered files"

The underlying computation and default parametes of some filters have changed.

[ckuenne]

Yes, can I send them here: michael.piechotta (at) informatik.hu-berlin.de? I do not want to make these public.

[piechottam]

Yes, BUT please replace the last part with "at gmail.com".

[ckuenne]

I just sent it. The attachment is 17MB zipped, so I hope gmail can deal with that.
...
[edited] wrong button..

[ckuenne]

I did some more digging: when looking only at the ~360k variants "lost" in Jacusa 2.0 when comparing vs. Jacusa 1.3 it comes all down to the strand.

minus strand: 360480
plus strand: 37

I then checked all variants for the strand.

1.3 minus strand: 440,549 
1.3 plus strand: 96,792
2.0 minus strand: 105,029
2.0 plus strand: 96,780

It lookes to me like Jacusa 1.3 simply had a bug that did skip certain filters for minus-strand variants.

[YiweiNiu]

Hello, is it a bug of JACUSA 1.3? @piechottam

[piechottam]

We generally recommend using JACUSA2. JACUSA1.x has been deprecated and won't be receiving any updates.
The underlying computation and default parametes of some filters have changed.

[YiweiNiu]

Thanks for your quick reply. Got it now.