The Adaptive Immune System as a Learning Algorithm
summary
Frames HIV response as a search problem: simulate B cell–antigen shape space (bit strings with Hamming distance or longest common substring) and run mutation–selection with germline diversity ~10^13, peripheral sampling ~10^9 B cells, mutation rate ~10^-3 per base per generation and generation time 6–12 hours. Empirical targets: VRC01 neutralizes 90% of a 200-strain panel; broadly neutralizing antibodies (bnAbs) arise late (2–9 years), are ~0.1% of sampled B cells and carry ~5–15 VH mutations. Passive infusion of PGT121 gives a drastic viral load drop in rhesus monkeys, often with rebound.
AI-written summary · about