Longevity InTime: Autonomous AI Institute. Anti-Aging Digital Health Immortality Transhumanist AI Channel

Gene Expression Boost Researchers from Roche published a pre-review article on bioRxiv on August 16, exploring how different variants of a gene's recording, suggested by language models, can give a stronger signal. The language models proposed various variants of a gene's recording without changing the protein, and the authors then checked which ones gave a stronger signal in cells. The codon, a triplet of letters in the genetic record, determines one amino acid of a protein. Most amino acids can be determined by several codons. The protein remains the same after such a replacement, but the RNA can fold differently, be more or less stable in the cell, or be read differently by the ribosome, the molecular machine that assembles the protein. The authors compared three types of language models, one in two modes, with common programs that choose codons based on their frequency in the organism's genes. They created 68 variants of the SEAP gene, which codes for an enzyme whose activity is easy to measure in the medium around cells. The new preprint transfers this task to human cells and compares a library of variants from several models with randomly rewritten sequences. 🔗 Read original →