search

PengNi / ccsmeth

Detecting DNA methylation from PacBio CCS reads
BSD 3-Clause Clear License
71 stars 10 forks source link

ccsmeth train #6

Closed SarahChen0401 closed 2 years ago

SarahChen0401 commented 2 years ago

Hi Peng, I would like to retrain your ccsmeth model using another ccs or clr dataset with corresponding WGBS data. I am not sure about your "train_file", is it an aligned bam file, feature .tsv file formatted by step 2 'ccsmeth extract'? How can I label the methylation level of every site or every ZMW? 

ccsmeth train -h
usage: ccsmeth train [-h] --train_file TRAIN_FILE --valid_file VALID_FILE
                     --model_dir MODEL_DIR
  --train_file TRAIN_FILE
  --valid_file VALID_FILE

Can you give me an example command? Thanks!

Best wishes, Shuhua

PengNi commented 2 years ago

Hi Shuhua,

You can refer this issue in deepsignal. ccsmeth has a similar training procedure. For training it needs feature.tsv files extracted using high-confidence methylatied/unmethylated (in my tests I used methylation frequency in BS = 1.0/0.0 ) sites.

Best, Peng