Protein language model
A transformer trained on amino acid sequences the way a text model trains on words — learning which residues tend to follow which, and what that implies about structure and function.
Learn first
A protein is a chain of amino acids, each one a letter in a 20-character alphabet. A protein language model trains on millions of known sequences without any structural labels, predicting masked or next residues the way a text model predicts masked words. In doing so it infers evolutionary grammar: which substitutions proteins tolerate, which positions are load-bearing. That internal representation turns out to encode three-dimensional structure and, to a degree, function — making the model useful for predicting how a mutation changes a protein, or for generating sequences with desired properties from scratch.
Where it came from
In megatrends
Related players
Finds citing this concept
Overall = breadth + depth + substance + ½·novelty + freshness (a small boost that halves each week)
The model rewrote a third of the protein
OpenAI's protein language model redesigned two of the factors that turn an adult cell back into a stem cell — and the winning variants differ from the natural proteins by more than a hundred amino acids.
How this connects
Tap a node to open it