f5c icon indicating copy to clipboard operation
f5c copied to clipboard

few sites from LSK114

Open ogrecio opened this issue 1 year ago • 3 comments

Hi, I'm using f5c to call methylation sites from LSK114.

I'm obtaining only few sites. Is f5c ready to be used on LSK114? How should I implement f5c on reads from LSK114?

Thank you, Oscar

ogrecio avatar Oct 25 '22 14:10 ogrecio

Unfortunately, we are still in the process of deriving a model for R10.4.1 (aka LSK 114). The current built-in models in f5c are for R9.4.1, so I am surprised it even picks a few sites in LSK114.

@hiruna72 is working on deriving a model - will let you know soon as we get it.

hasindu2008 avatar Oct 25 '22 14:10 hasindu2008

Thank you for the reply. Looking forward to it. Any tentative date? :-)

ogrecio avatar Oct 25 '22 15:10 ogrecio

If it goes according to the plan, like one month. But these things rarely go according to the plan :D. So I would say a couple of months.

hasindu2008 avatar Oct 26 '22 10:10 hasindu2008

@ogrecio Sorry for the delay. Finally, we have a very crude methylation model for R10.4.1. Do you by chance have some data with a baseline suitable for testing?

hasindu2008 avatar Mar 04 '23 00:03 hasindu2008

Hi! we may have some data available. Maybe you can contact me privately and give me some details on what you would need.

Cheers, Oscar

ogrecio avatar Mar 04 '23 14:03 ogrecio

On Twitter? Seems I cannot reach you there, possibly due to a setting. Could do on email if ou provide.

Also, if you are keen to try running, you can try the following steps:

git clone https://github.com/hasindu2008/f5c/ -b r10
autoreconf             
scripts/install-hts.sh  
./configure
make 

Then, run the test script for r10:

scripts/test_lsk114.sh
head test/hg2_lsk114_reads_1000/meth.tsv # to see if there are any calls

Then, you should specify the options models manually using options --kmer-model test/r10-models/r10.4.1_400bps.nucleotide.9mer.template.model --meth-model test/r10-models/r10.4.1_400bps.cpg.9mer.template.model to f5c.

hasindu2008 avatar Mar 06 '23 04:03 hasindu2008

@ogrecio

I have done a prerelease f5c-v1.2-beta with the r10 models inbuilt. Tested on some human data and got ~0.9 correlation with bisulphite data (without filtering). Great if you can run on your samples and check if you get more methylation sites.

hasindu2008 avatar Apr 04 '23 00:04 hasindu2008