Here a question for immuno-informaticians among us (hope they are here at all):
We are attempting a reverse vaccinology approach on a fish-parasite (non-)model (Salmo salar and Lepeophtheirus salmonis). There we try to predict candidates from the proteome that would give good vaccine candidates. We found a pipeline, called vacceed, that might work for us. Reverse vaccinology approaches include predicting binding affinities of peptides or epitopes to the MHC-I and MHC-II classes using tools like IEDB. These tools have been mainly used sucessfully on mammalian data in my understanding, and therefore I have assumed so far that the results would not transfer well to a fish. Even in humans, the various different MHC alleles give very different binding affinities, and therefore MHC-I/II predictors (like IEDB, we intend to use the stand-alone Linux version) need the present alleles as input files. For salmon, i found a review on evolution of MHC alleles in teleost fish, but I am not quite sure how to prepare allele files from the presented information. Therefore my questions:
- How to prepare allele information for a fish as input to IEDB MHC predictors in the proper format?
- Will the result be any better than random? (guess this is hard to determine)
Example of an MHC-I allele file for cattle (from the vacceed manual):
BoLA-T2C,10
BoLA-T2C,11
BoLA-T2C,12
BoLA-T2C,13
BoLA-T2C,14
BoLA-T2C,8
BoLA-T2C,9
Example of an MHC-II allele file for humans
H2-IAb
H2-IAd
H2-IAs
H2-IEd
HLA-DPA1*01/DPB1*04:01
HLA-DPA1*01:03/DPB1*02:01
There are no prediction methods in that IEDB tool that allow for predictions for fish, so why do you want to use it again?
In any case, you did not link the 'presented alleles' from the paper (and the reference in the paper doesn't work, I think). Do you mean these alleles? : https://www.ebi.ac.uk/ipd/mhc/group/FISH/allele?search=sasa&submit=1 You could either scrape that page, or just download the result, and parse the fasta file?
Hi, and thank you very much for the link. For explanation, L. salmonis is an ectoparasite of salmon. I want to use the standalone version of the IEDB MHC tools from http://tools.iedb.org/mhci/download/ I think the tools have a configuration file to specify the alleles present in the organism. So in principle, now I could scrape the Sasa-alleles for that config file, however I am not sure that this makes sense, as these alleles have not been part of the training set. I am assuming that this does not make much sense overall, on the other hand, there is the strong wish to predict some sort of candidates, so I rather need a good argument why it won't work.
Hi Michael
There exist some tools that try to predict binding affinity if you feed protein sequences aswell, but unless the binding affinity for fish (your salmonā¾ have been taken up in an existing predicting tool you're out of luck. You could start making your own tool if there are even binding affinities available for the salmon, and your last resort should be the predicting tool that used protein sequences I guess?
Hi Mathias,
I don't think the MHC affinities have been measured for the salmon, in my understanding this can be done experimentally but is very costly. At least searches for "mhc binding affinity" +salmon or t-cell epitope prediction +salmon did not turn up with something significant, except for that one of the prominent author's surname in that field is Salmon. So if I understand correctly, if someone wanted reliable epitope predictions for this organism, they should first invest into laying the proper foundation by MHC affinity measurements, or what did you mean with that "last resort" tool, which tool are you referring to?
Btw, it seems to be a pattern that stakeholders want some sort predictions, but are not willing to invest into proper data.
Yeah if you want reliable predictions, you'd need binding affinities of the salmon alleles.
But I was told there exist other prediction tools that predict binding affinities based on the protein sequence of your allele. So you could download the protein FASTA on ipd-mhc db and feed it to these tools. These tools should be your last resort, only if tools based on actual binding affinities don't exist - which is your case.
I don't know any of these tools though, you'll have to look for them, if they indeed exist :/ (I worked with netMHCpan last year for a small project, but that's about it when it comes to the immuno-info strategies I'm afraid).
Btw, I was thinking to cross-post this on bioinformatics.SE, any objections?