3 ms·
There's some good stuff in this but while I agree with many of his points, I don't agree with his fundamental conclusion. I see several examples of "machines"
by dekhn 4y ago
There's some good stuff in this but while I agree with many of his points, I don't agree with his fundamental conclusion. I see several examples of "machines" in biology which satisfy all my requirements. In particular DNA and RNA transcriptase, the ribosome, and motor proteins are all machines. They act with very high fidelity, have a specific and necessary purpose (to the extent that an evolved system can have a well-defined purpose), consume energy to achieve their goals, and eventually break down and make mistakes.
The current belief about alphafold is that its confidence is actually a "disorder predictor": when it makes low confidence predictions, it's really that the protein itself doesn't adopt a well-defined structure.
But further than this, we actually do protein design and create new proteins that have well-defined functions. FOr example there is a class of proteins known as proteases that "cut proteins" (conceptually like scissors, but physically using enzymatic activity). A well-known protein designer at Genentech took an existing protease and made it more heat-stable (IE, didn't stop working when you raised the temperature). This protein was then sold licensed to laundry manufacturers who added it to their detergent (it does a great job cleaning up protein stains, hence the advertisement "protein gets out protein"). It's literally a machine you put in your machine to make your machine machine better.