search

weir12 / DENA

Deep learning model used to detect RNA m6a with read level based on the Nanopore direct RNA data.
MIT License
22 stars 5 forks source link

Regarding about the output of LSTM_predict.py. #17

Closed guowb9 closed 1 year ago

guowb9 commented 1 year ago

Hi, I have used the DENA tools to detect RNA m6A in my data, and the "-d" parameter was added when I ran LSTM_predict.py. I got result ${prefix_outfile}_details.tsv like this:

ENST00000000233.9 317 AAACA df2e44fd-337f-4d04-bcd6-95d9ce62e6ef 1.4540097 21a9b4c5-8ab5-42c2-a561-4aa9fffc6688 -2.8870084 dd4e2a5d-dfbe-45fd-bd18-c4b15ae12bae -0.56397027 a68bbb3e-93c4-4162-beeb-e1620273977f 0.92993855 6be8d8b0-b897-47ba-a86d-ff83b3e02ea4 0.026935648

As your description: the first column was the read ID; Second column was the m6A-modified probability of this read at the candidate coordinate on ENST00000000233.9). But I find the m6A-modified probability <0 and even >1, is the output result right? I didn't meet errors when run the process. Thanks!

weir12 commented 1 year ago

Hi, Please check whether the model file name extension is .dat

The .dat file triggers the softmax function of pytorch in the if branch. (This step is not required for other machine learning models that we have been tried in the past.)

https://github.com/weir12/DENA/blob/2c584c9a22f2903a1c44abe9a734fef5e9d158c8/step4_predict/LSTM_predict.py#L97

Best wishes