I went down the neoantigen rabbithole, and it was quite interesting.
I liked this talk on “Developing Personalized Neoantigen-Based Cancer Vaccines”.
It seems a core part of their methodology is using machine learning to predict which peptides will elicit a T-cell response, based on sequencing the patient’s tumour. (Discussed starting from around 11 minutes in.)
They use this algorithm, which seems to be a neural network with a single hidden layer just ~60 neurons wide, and some amount of handcrafting of input features (based on papers from 2003 and 2009). I wonder what one could accomplish with more modern tools (though I haven’t yet read the papers deeply enough to have a model of how big of a bottleneck this is to creating an effective treatment, and how much room for improvement there is).
They use this algorithm, which seems to be a neural network with a single hidden layer just ~60 neurons wide, and some amount of handcrafting of input features (based on papers from 2003 and 2009).
To me the website looks like the published in 2020 a paper with the newest version. I would expect that you can solve any problems with that algorithm underperforming by simply taking more plausible neoantigens. If you want 20 one’s that actually work and you have a successrate of 50% you can just take 40 different peptides. That makes it a bit more expensive but still doable.
I would expect that in the span of this decade AlphaFold gets the capability to model those interactions nearly perfectly but without large resources I don’t think you will easily improve on the existing bioinformatics models.
I went down the neoantigen rabbithole, and it was quite interesting.
I liked this talk on “Developing Personalized Neoantigen-Based Cancer Vaccines”.
It seems a core part of their methodology is using machine learning to predict which peptides will elicit a T-cell response, based on sequencing the patient’s tumour. (Discussed starting from around 11 minutes in.)
They use this algorithm, which seems to be a neural network with a single hidden layer just ~60 neurons wide, and some amount of handcrafting of input features (based on papers from 2003 and 2009). I wonder what one could accomplish with more modern tools (though I haven’t yet read the papers deeply enough to have a model of how big of a bottleneck this is to creating an effective treatment, and how much room for improvement there is).
To me the website looks like the published in 2020 a paper with the newest version. I would expect that you can solve any problems with that algorithm underperforming by simply taking more plausible neoantigens. If you want 20 one’s that actually work and you have a successrate of 50% you can just take 40 different peptides. That makes it a bit more expensive but still doable.
I would expect that in the span of this decade AlphaFold gets the capability to model those interactions nearly perfectly but without large resources I don’t think you will easily improve on the existing bioinformatics models.