search

Shamir-Lab / DOMINO

Network-based module discovery algorithm with high rate of empirically-validated term calls
MIT License
9 stars 2 forks source link

output differences #5

Closed BetulCeylan closed 2 years ago

BetulCeylan commented 2 years ago

Hi!

I run DOMINO consecutively with the same imputs (same network and active genes list) and get different module numbers and sizes. What is the reason behind of this situation? When I check the intersection of union of these modules that are produced by seperate run , they have ~80% common genes but the rest of them are different. Which result can we choose or trust in this case?

Thanks, Betul

hag007 commented 2 years ago

Hi Betul,

As DOMINO has probablistic components in it (i.e the first dissection of the network, the PCST step), minor differences among executions are expected. Therefore, it is possible to get ~80% mutual genes in some settings (depends on the set of active genes and the network).

Nevertheless, results of different executions usually reflects a similar biological picture (e.g. similar GO term enrichment scores).

Best, Hagai

BetulCeylan commented 2 years ago

Thanks for explanation!