Prediction: Chromosomes do not match between input files. about jstacs HOT 10 CLOSED

Hi,

my best guess would be that your FAI file does contain a chromosome that is not present in the input files (chromatin accessibility and motif files). When predicting, you may specify "prediction chromosomes" with parameter p, and if that is not present, the prediction chromosomes are internally set to all chromosome present in the FAI (in sorted order). The process then starts looking for the current chromosome by sequentially scanning all input files line by line and if that chromosome cannot be found at any line, the readers scanning those input files finally return "null", which might be the case, here.
An alternative explanation might be that all input files do not contain a single line (and the readers returning "null" right from the beginning), but as those files worked in the training step, I would consider this rather unlikely.
If that does not explain your issue, I would provide a Catchitt version with additional debugging output that could be posted here to localise the error.

Best,

Jan

from jstacs.

Thanks for the quick reply!

I have tried predicting using individual chromosomes and found chrY cause this error. But chrY is present in both Motif and Accessibility features.tsv (and in .fai file). I don't know the reason for my bug but I basically get what I need except for chrY, which is really not a problem for me.

Thanks,
Yichao

from jstacs.

Thanks for the swift reply and great that you found a way working around this issue.
However, I would appreciate if you could check that the IDs from "chrY" are exactly identical (including upper/lower case) in the FAI and feature files? If yes, there might be a bug on our side (although I currently don't see where this might come from), and it would be important for us to correct it.

Best,
Jan

from jstacs.

Yes, they are exactly the same, "chrY".

Thanks,
Yichao

from jstacs.

Would you mind posting your FAI file and the file output of
gunzip -c Chromatin_accessibility.tsv.gz | grep "chrY" | head -n 1000 | gzip > test_access.tsv.gz
of your accessiblity file and, in analogy,
gunzip -c Motif_scores.tsv.gz | grep "chrY" | head -n 1000 | gzip > test_motif.tsv.gz
for one of your motif files?

from jstacs.

Sorry for my late response.

[yli11@noderome105 fasta]$ cat mm9_main.fa.fai 
chr1	197195432	6	50	51
chr10	129993255	201139354	50	51
chr11	121843856	333732482	50	51
chr12	121257530	458013223	50	51
chr13	120284312	581695911	50	51
chr14	125194864	704385917	50	51
chr15	103494974	832084686	50	51
chr16	98319150	937649567	50	51
chr17	95272651	1037935107	50	51
chr18	90772031	1135113219	50	51
chr19	61342430	1227700698	50	51
chr2	181748087	1290269983	50	51
chr3	159599783	1475653038	50	51
chr4	155630120	1638444823	50	51
chr5	152537259	1797187552	50	51
chr6	149517037	1952775563	50	51
chr7	152524553	2105282947	50	51
chr8	131738871	2260857998	50	51
chr9	124076172	2395231653	50	51
chrM	16299	2521789355	50	51
chrX	166650296	2521805986	50	51
chrY	15902555	2691789294	50	51

test_motif.tsv.gz

test_access.tsv.gz

from jstacs.

The problem seems to occur if a) the chromosomes do not appear in alphabetical order in the motif/accessibility files and the FeatureReader needs to re-start to search for the chromosome from the beginning. Should be fixed with commit da23fb1

from jstacs.

Great! If you can provide a jar file, then I can quickly test it out.

I'm not good at Java, not sure how to get a jar file from .java file.

Thanks,
Yichao

from jstacs.

Please find the JAR at https://cloud.uzi.uni-halle.de/owncloud/index.php/s/hLgqnVomv3ZGzkR
If it works for you, I would make this a public bugfix release.

from jstacs.

It worked!

Thanks,
Yichao

from jstacs.