Hello, I have passed some soil samples through funscan for AMP discovery. I have got the Ampcombi.tsv but not sure how to interpret the data. For example out of 837423 protein sequences, the AMP score is 1 for over 250,000 of them using AMPlify, for Ampir only 7 sequences have a score of 1, and for Macrel there are no sequences with a score of 1. Should I only look at ones that have pass a particular threshold for all 3 methods for example? Any advice would be greatly appreciated. Thanks.

That you will need to ask @Anan Ibrahim about, I'm not sure what the AMPcombi score refers to.

In that case, if each AMP tool does things a bit differently so the output can mean slightly different things, then what's even the point of merging their results, if they can't be properly compared across the different tools?

Sorry to clarify: I mean the way they find AMP hits are different

Thanks @James Fellows Yates, makes sense. If you run AMPlify manually, you can train it with a more imbalanced dataset and that reduced my proportions of predicted AMPs significantly. I couldn't find such a setting on the funcscan. Is that a possible added functionality for the future?

@Nick Eckersley you should be able to add custom arguments to AMPlify via

ext.args

in your custom config

It depends how complicated the retraining is @Nick Eckersley - if it's a simply flag (can is testable with the suggestion from @Alexandru Mizeranschi), then yes that woudl be easily implementable