Pinecone is our public protein design software. Feel free to play around. Feedback is welcomed. We're trying to build an abstraction layer above DNA sequences so you can build genetic tools to fulfill particular capabilities without needing to know how to build DNA.

Or just to help teach what useful designs others have put out there - reveal some of the 'magic' of current gene therapies or genetic editing tools. Some interesting examples: https://serotiny.bio/notes/proteins/

Neat! What sorts of design decisions did you have to confront when designing this language? Have you found it useful as a toy modelling space, i.e. where suitably replicating real-world protein X in “code” reveals interesting things about X?

A bit of a late reply, sorry, but it's a great question.

We found many design decisions we had to confront - more than we expected. The largest two include 1) the decision to use an amino acid sequence as ground truth (and not DNA), and 2) the decision to make the language universally proscriptive (at the cost of being universally descriptive).

Anything you build in the system can be altered, broken apart, and itself be part of a larger whole - all while being manufacturable. But that comes at the cost that there are certain highly complicated overlapping synergistic type designs that just can't be made. Whereas with an entirely descriptive language one can describe the multi-layered fugue that is the HIV genome - but the construction rules become so lax that they lose their instructive capabilities, and can no longer logically guarantee constructibility.

The most interesting (and surprising?) thing we learned is that we now have the only data-structure that can reliable and efficiently answer questions of the sort of, "show me any/all designs that utilize the protein, _GFP_".