search

alyosama / virnet

VirNet: A deep attention model for viral reads identification
Apache License 2.0
30 stars 4 forks source link

work on contigs (2.5kbup); result #2

Closed Ruonan0101 closed 4 years ago

Ruonan0101 commented 5 years ago

Dear developers, I have two quick questions. 1) will virnet work on the contigs with length greater than 2.5kb by using --input_dim=3000; 2) after running on my test file, I have problem to interperate the results. what 'score' in the 4th column refers to and what does the '1' and '0' in result mean. I assume it is binary , present or not and want to confirm with you. Thanks~

alyosama commented 5 years ago

Dear @Ruonan0101,

  1. we didn't use it 3000 model with 2.5kb before but I think it will work. Let me know what will you find. Instead you can make all contigs with 1000kb and run the tool with --input_dim=1000.

  2. The score is the classifier probability of classify the read to viral or not viral. (1 is viral and 0 is not viral(bacteria or archaea)

Let me know if you need anything else.