Blog
Longer reads from the cancer3.ai team — explained carefully, illustrated, and sourced.
Thirty-four letters of the lock: how antigen-presentation models actually work
Our previous article gave antigen-presentation models five paragraphs. Here we take them apart properly — and from scratch, assuming no biology. What an allele is and why the same cell looks different to two people, where the idea of describing the HLA groove with thirty-four letters comes from, how the mass-spectrometry data is physically produced, why the model invents its own missing labels, and what the numbers these tools advertise actually measure.
Read the article →
The algorithms behind the vaccine: which machine-learning models pick a tumour's 34 targets
An individualized mRNA vaccine targets 34 neoantigens chosen from thousands of mutations. That choice is not made by a human — it is made by neural networks. We take the pipeline apart step by step: variant calling, pan-specific antigen-presentation models, deep learning on mass spectrometry — and we say plainly where public knowledge ends and proprietary code begins.
Read the article →This page is educational — it is not medical advice and does not replace consultation with an oncologist. Diagnostic and treatment decisions are made solely by specialist physicians.