Classes II ?
zimoun opened this issue · comments
Dear main contributors @andras86 @b-schubert
Thank you to release this open-source efficient code.
From my few experiments, it works well for Class I. Nice! :-)
Obviously, more is better :-) and Class II should be nicer.
From what I have quickly read in your paper and in your code, your proposed strategy should work (more or less) with any Class. Right ?
Less in the sense that the Integer Linear Program will be harder and harder to solve when increasing the number of alleles, since the matrix of constraints will be large (or extremely large).
Therefore, I would like asking:
-
Is it still planned to release an updated version ?
- If yes, is it available elsewhere ? branch, repo, etc. Even an half-cooked version
- If no, does it exist documented attempts ? what passed and what failed ? Saving some time to avoid already known traps.
All the best
Hi zimoun, it's in the works. It's been way longer than it should have been, but yes, it will get released. No half-cooked version will be made available though, sorry.
Thank you for your quick answer.
Since your answer about point 1. is yes,
- Do you have a time-schedule about this release of Class II ?
- Still open-source ?
- And I am guessing that you are waiting the publication of the companion paper before releasing, right ?
There is a new comparison of HLA typing algorithms published. It reports that OptiType is the only algorithm which can't make class II allele predictions. It means that you could use any competitor's software to get such predictions.
@DarioS : Thank you for the pointer.
I already knew some typers presented in, tested some of them, but I was not aware of all. I will try with my data.
However, this does not answer to my point initial open issue.
OptiType formulates the problem with the lens of mathematical optimisation. And it is demonstrated efficient for the Class I.
The question/issue is: does the same algorithm based on Integer Linear Program still work for the Class II ?
From my knowledge, there is no answer yet because no one has implemented the algorithm for the Class II. I mean, nothing publicly released. Therefore, I was asking if it was implemented.
Your claim "OptiType is the only algorithm which can't make class II allele predictions" is not accurate enough. The current implementation cannot make Class II allele prediction, but it does not appear clear to me if the algorithm can or cannot predict the Class II.
Since it is almost one year, maybe the authors @andras86 @b-schubert are not interested anymore. Well it is open source, I should implement myself to answer my points :-)
Thank you again for sharing the pointer.
All the best !
The algorithm itself can predict Class II, but as it turned out requires major adaptations to do it accurately enough for our standards. The release will happen eventually, the project isn't dead.
@andras86 Did you implement these major adaptations ?
Hum? the "straightforward" extend is it not enough accurate ?
Where does come from the major issue ?
- from the way to write the system to optimise ?
- or from the ILP solver ? (solution not global etc.)
All the best !
Hi @andras86,
Nearly 2 years down the line since your last update here - can you let us know whether you're still planning a release for Class II typing? I only ask because there's recent activity on the repository, I'm hoping it may be picked up again.
Thanks
Yes, it will finally arrive soon.
Any news about Class II?
Any news about Class II? Thanks
Same question!
Check out the hla-ligand-atlas branch for now. Even with its limitations, the early feedback is positive.
Check out the hla-ligand-atlas branch for now. Even with its limitations, the early feedback is positive.
Hi, @andras86
i am so appreciated that you develop such wonderfull hla-typing tools. I have used hla-ligand-atlas branch to analysis in 30 samples. it perform excellent. However, i want to know what release of hla-imgt do you use in the classII branch.