how to use conservation flag

[Yogesh1-11]

HI, i saw a conservation flag in prankweb
Use original structure
Use conservation

but in local readme there is no instruction of selecting conservation option

[davidhoksza]

Hi, the command-line application does not support conservation directly. This is because calculating it requires a 10GB dataset, so it was developed separately from P2rank and is only used in the web application.

That said, you can compute conservation yourself and use it as described here. Another option is to use p2rank framework with Docker containers.

[Yogesh1-11]

really? i was able to reduce conservation groovy config to reduce the number of pocket. isnt it for same?

…

On Thu, Jan 29, 2026 at 5:19 AM David Hoksza ***@***.***> wrote: *davidhoksza* left a comment (rdk/p2rank#81) <#81 (comment)> Hi, the command line application does not support conservation. This is because calculating it requires a 10GB dataset, so it was developed separately from P2rank and is only used in the web application. — Reply to this email directly, view it on GitHub <#81 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AZNDTOU7MNB2YSFZKISWMHT4JERVLAVCNFSM6AAAAACTFWW3AWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTQMJTHE4DMNZSGQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[rdk]

Hi @Yogesh1-11, thank you for your interest in P2Rank.

Running P2Rank on the command line with conservation (which corresponds to using PrankWeb with the conservation option) is not currently supported or documented. To use it meaningfully, you need to calculate residue conservation scores first - and there's no easy, well-documented way to do this currently.

That said, it is implemented, so you can technically run P2Rank with -c conservation_hmm.groovy, but it doesn't make much sense without providing calculated conservations via the -conservation_dirs {dir_with_precalculated_conservations} param. If you run it without providing conservation scores, any reduction in the number of pockets is coincidental and the results will generally be worse - you're not actually benefiting from the conservation models. If you just need to reduce the number of output pockets, there are better ways to do it.

Could you share more about your use case? Specifically:

• Do you just need to reduce the number of predicted pockets?
• Are you trying to run conservation models on a fixed dataset for benchmarking?
• Or do you need this to work for any given protein as part of a pipeline?