Problem about Deepsignal in 6mA detection.

bioinfomaticsCSU / deepsignal

Detecting methylation using signal-level features from Nanopore sequencing reads

GNU General Public License v3.0

109 stars 21 forks source link

Problem about Deepsignal in 6mA detection. #82

Open Bin-Ma opened 2 years ago

Bin-Ma commented 2 years ago

Hi, @PengNi I tried the Deepsignal for 6mA detection and all the position is matched with the 5mC modification on the reference sequence. I don't know what happened about my command. Looking forward to your reply.

deepsignal call_mods --motifs GATC --input_path ./barcode01/deepsignal_feature.tsv --is_gpu yes --model_path ~/software/deepsignal/model.GATC.R9_2D.tem.puc19.bn17.sn360/bn_17.sn_360.epoch_5.ckpt --result_file ./barcode01/deepsignal_prob_A.tsv

PengNi commented 2 years ago

Hi @Bin-Ma , in your step deepsignal extract to generate deepsignal_feature.tsv, the --motifs should be set as --motifs GATC.

Best, Peng

wenluo711 commented 2 years ago

Hi @PengNi ,

I'm also trying to detect 6mA and 5mC in bacterial sample but only find the model model.GATC.R9_2D.tem.puc19.bn17.sn360/bn_17.sn_360.epoch_5.ckpt in the google drive that are supposed to using on version <0.16. For Deepsignal2, there's only one human CG motif. Does that mean users has to train their model by themselves in such cases? Also for different motif patterns do we all have to train a separate model for each motif?

Thanks, Wen

PengNi commented 2 years ago

@wenluo711 , if you want to detect 6mA and 5mC in bacterial samples, I suggest using Remora/mCaller/nanodisco. Right now deepsignal/deepsignal2 only performs well in 5mCpG detection.

Best, Peng

Bin-Ma commented 1 year ago

@PengNi Thanks for your reply. I finished the 6mA detection using Deepsignal now. However, the correlation against other software (DeepMod, mCaller) was low. I don't know how to set the cutoff value of Deepsignal. Looking forward to your reply.

Best, Bin