Puumanamana / CoCoNet

Tool for unsupervised contig binning from viral metagenomes
Apache License 2.0
15 stars 1 forks source link

How to pre-cluster? #4

Open Yangzyangjin opened 2 years ago

Yangzyangjin commented 2 years ago

Dear CoCoNet developer,

I am really interested in your tool. I read your paper,and paid more attention to your clutering algorithm. iIn the step 1 "Edge computation ", image , i don't know how to caculate the probalitities of contig v_i,v_j? Is there any code?

Thanks!

Best, Herina Yang

Puumanamana commented 2 years ago

Hi Herina, Why are you trying to compute those probabilities? You can access the number of expected hits if you load the "graph-*.pkl" file (after running coconet) and check the weight of the edge between two contigs. Cédric

Yangzyangjin commented 2 years ago

Hi Cédric, I want to apply the edge computation method to the new dataset,and i want to reproduct your clustering algorithm,but i don't know how to caculate the probabilities?

Puumanamana commented 2 years ago

Well, as I said, you can directly access the expected hits from the graph file. If you want the detail of all the individual probabilities for each pair of fragments between individual contigs, that is going to be more complicated (it requires loading the neural network model and the weights, loading the features for all contigs, and then running the network for a given contig pair).

Yangzyangjin commented 2 years ago

well,i think i understand what you mean.Thanks for your patience! Lastly,what is the formula of P ?Is this a distance metric?

Puumanamana commented 2 years ago

P is the probability returned by the neural network