Identifying Multiple Domains Using Hmmer 3.0

Question

[Hari]

Hi all, I used HMMER 3 to identify DH (Rhogef) domains related to human .So far i am able to extract all the DH domains.However i would like to know how i can find all the domains that occur in combination with DH ex.DH-PH or DH-CNH,PFAM shows me http://pfam.sanger.ac.uk/family/PF00621#tabview=tab1 that there are around 392 possible RHogef combinations.So how do i construct a profile HMM so that i can mine all the DH domains that occur with other domains.I tried to download the DH domain(rho gef) sequences that are in PFAM to construct my profile HMM,but this finds only the DH domain but cannot mine the other domains.How do i approach this?

Thanks in advance

Answer 1

[Michael Kuhn]

So how do i construct a profile HMM so that i can mine all the DH domains that occur with other domains.

An HMM profile only works well for one domain, and you need to use different HMM profiles for other domains. (If you build a combined profile, and the order of the domains changes or another domain appears in the middle, you won't get a hit to the full combined profile.)

To find out which domains co-occur with DH, you need to download / build profiles for all the other domains, and then search for these domains in the proteins which you identified to contain DH.

[Hari]

dh domain has too many co-occuring domains,to build a hhmm profile for all of them will be too painful.thanks for the suggestions.

Answer 2

[Pablacious]

However i would like to know how i can find all the domains that occur in combination with DH ex.DH-PH or DH-CNHcombination with DH ex.DH-PH or DH-CNH

For this particular part of the question, I would search for the PFam or Interpro domain in UniProt (in the main search box, just paste the id).

You get approximately 5,391 proteins for PF00621 (you can further limit there by taxonomy, for the organism or taxa you are interested). Then where it says Results, there is a Customize button which you could use to display only the InterPro domains, which for this search are displayed by default (but then getting rid of other fields might make the download smaller later). Independently of the previous step, go to download, choose your favourite format, and you will get a file that lists all domains (using InterPro IDs, but I'm sure this can be easily converted to PFAM ids) that are occurring with your domain of interest.

With a bit of filtering (I'm thinking probably awk and sort -u if you downloaded a tab-separated text file) you can easily get from here to a unique list of domains. Then you can download those domain's HMMERs and do your search (with each one of them, as Michael correctly pointed out, there would be no point in merging to HMMER models in the same model).

Hope it helps!

[Hari]

thanks for the suggestions.

[Hari]

+1 for the suggestions,thank you

Answer 3

[Hari]

Hi all,

I was able to detect mutiple domains with hmmscan,all i downloaded the whole Pfam-A.hmm (profile hmm) from pfam and gave my query hmmscan does it automagically.

Answer 4

[mlwbglorious]

Hi Hari, I am not sure which file is the one you mentioned, could you maybe send me a link ? Many thanks, that'll be be very handy to download all the hmms.