Sadly, entirely predictable.
-
And why the journal pushed it through so quickly, wrote an editorial on it and hawked it to the world’s press.
I have done these types of analyses - looking at how well codon usages of viruses match those of (known) hosts to train a model to predict putative hosts for novel pathogens. Good idea, but it can’t be done - rarely is the top match a know host and there is no meaningful signal!
-
-
Out of curiosity - given that ML-based projects on this especially aren't likely to slow down in the next 10 years, what do you think good work on this would look like? (I realize this might be a conversation for over beers instead of twitter)
-
Oh yes, AI to the rescue
. Issue one - small training dataset and that's hard to overcome. But otherwise, integration of other data - trees, kmer profiling, restriction sites, host gene interactions, structures, known receptors, etc. - Nog 3 antwoorden
Nieuw gesprek -
-
-
I specifically stopped myself from retweeting this article yesterday and was surprised to see how quickly this spread
-
I think its still an article and its spread has generated very useful discussion that may not have happened otherwise.
- Nog 2 antwoorden
Nieuw gesprek -
Het laden lijkt wat langer te duren.
Twitter is mogelijk overbelast of ondervindt een tijdelijke onderbreking. Probeer het opnieuw of bekijk de Twitter-status voor meer informatie.