search

cdanielmachado / carveme

CarveMe: genome-scale metabolic model reconstruction
Other
149 stars 51 forks source link

Cut-off to choose the homology template using DIAMOND #22

Closed hongzhonglu closed 6 years ago

hongzhonglu commented 6 years ago

Hi Daniel, Great work! I have a question about how to set the cut-off values in the blast analysis using DIAMOND. Looking forward to your reply. Thanks in advance!

cdanielmachado commented 6 years ago

Hi @hongzhonglu

You can see in this file how we invoke diamond:

https://github.com/cdanielmachado/carveme/blob/master/carveme/reconstruction/diamond.py

We are currently running diamond with these options: "--more-sensitive --top 10", which takes the top hits instead of using an arbitrary cut-off value.

You can try passing any arguments you want to diamond with the flag --diamond-args and see if other types of cutoff improve your annotations.

We tried with many different options, but this was what worked best for us. Since CarveMe uses the homology scores to assign reaction scores, the low homology genes generate small reaction scores, so they do not influence the final reconstruction so much.

hongzhonglu commented 6 years ago

@cdanielmachado Great thanks!

hongzhonglu commented 6 years ago

Hi @cdanielmachado
An another question. Do the homology scores mean the bitscores, which is obtained using diamond? Thanks in advance!

cdanielmachado commented 6 years ago

Yes, the homology scores are the bitscores, sorry for not making that more clear on the paper.

hongzhonglu commented 6 years ago

Thanks very much!