Biology Isn't Just a Data Problem: The lncRNA ANRIL Breaks Every Model You'd Write
iskander · x · 2026-10-31
A long thread dismantling the "biology is just a data problem" premise with two real cases. GWAS found a strong heart-disease risk locus at 9p21 — a gene desert containing only the lncRNA ANRIL: primate-specific, 20 exons, dozens of isoforms, with linear and circular forms doing opposite things (linear recruits Polycomb to silence neighboring p15/p16; circular binds PES1 and jams ribosome assembly to kill overgrowing plaque cells). The risk allele shifts the linear/circular ratio, and the locus links to diabetes, aneurysms, glaucoma, gum disease and cancer. The quoted post adds humanin: a peptide encoded inside mitochondrial 16S rRNA that is simultaneously a ribosome structural component and a secreted anti-apoptotic signal — 24 aa if translated in the cytosol, 21 aa in mitochondria (different genetic code), plus 13 nuclear copies, and nobody is sure which source actually matters. Conclusion: whole-cell models keep hitting hierarchical, pleiotropic complexity that pure data-scaling doesn't capture.
More from AGI Musings
- Logic reasoning and neural nets, once dismissed, became AI's most used tools — luislamb · 2026-10-31
- Zhang Yiming's decade-old interview resurfaces as he tops China's rich list again — oran_ge · 2026-10-31
- Dev questions AI ethics: if education can't make humans moral, why trust aligned AI — luislamb · 2026-10-31
- The smartest people are splitting into two opposite paths: AI frontiers vs. inner retreat — oggyxe · 2026-10-31
- Robotics startups must deploy ASAP — sim and world models are just proxies, argues Wayve VP — m_wulfmeier · 2026-10-31
- Obesity drugs signal medicine's shift from treating to preventing disease — SydSteyerhart · 2026-10-31