search

openvax / mhcflurry

Peptide-MHC I binding affinity prediction
http://openvax.github.io/mhcflurry/
Apache License 2.0
193 stars 58 forks source link

is mhcflurry more accurate than NetMHCpan-4.0 #134

Closed kobejamescurry closed 5 years ago

kobejamescurry commented 5 years ago

Dear professor, in your outstanding paper, you mentioned that image so does the abstract mean there is no better performance of mhcflurry than NetMHCpan-4.0 in accuracy excluding speed. thanks a lot.

timodonnell commented 5 years ago

Hi there - this is correct. We did not see a statistically significant accuracy improvement over NetMHCpan 4.0 on our small validation set. Larger validation datasets would be needed to see an accuracy difference between the two (if there in fact is any).

kobejamescurry commented 5 years ago

Hi there - this is correct. We did not see a statistically significant accuracy improvement over NetMHCpan 4.0 on our small validation set. Larger validation datasets would be needed to see an accuracy difference between the two (if there in fact is any).

Thanks a lot. do you some suggestion on how to cut into peptides, I have not seen some papers referring to this. Here I will describe my concer more clear. after we get the vcf, we can download the reference of pep and cdna from ensemble. but how can I commbine the three data to get the 8-15mer peptides inclhding the mutant base. thanks a lot.

timodonnell commented 5 years ago

Hey, if you have RNA also (which is used to pick the transcript for each mutation), you could try using isovar or, for the particular case of making a synthetic long peptide vaccine, vaxrank. Both of these tools can use various MHC binding predictors, including NetMHC and MHCflurry. There are a number of tools developed by other groups for doing this, such as mupexi and pvacseq. Hope this helps!

kobejamescurry commented 5 years ago

Thanks a lot. your answer is very useful, I am using mupexi, which the result is not easy to understand. mine has just only dna-seq