Protean Language Model
This project is in development for the deCYPher residency, an interdisciplinary research program where machine learning is used to design protein sequences. In simplified terms, a protein language model generates candidate sequences that mirror a plant of interest, the sequence is engineered into a microbial host, and through precision fermentation the microbe produces a targeted plant enzyme or byproduct.
This project, Protean Language Model, is a pun on that “protein language model.” Protean means dynamic, shape-shifting, and it’s my playful way of negotiating a very complex pipeline and emphasizing its transformations.
I translate the deCYPher framework into four agents: Plant, Microbe, Machine, Human. Each agent stands for a component of the pipeline, represented by a specific figure.
- Plant: immortelle, a flowering plant associated with cell regeneration and cosmetic promises of renewal (including products like L’Occitane’s “Immortelle” line).
- Microbe: the soil bacterium used in deCYPher, chosen for its carotenoid metabolic pathway that can be co-opted to produce the desired compounds.
- Machine: the ML models used in the research workflow (including EMS-2).
- Human: the story engine, anchored in the long pursuit of immortality, from Qin Shi Huang’s obsession with eternal life to contemporary tech funded longevity research.
These four agents form an ecosystem inside this model, generating musical chords through mapped conditions. Ultimately, the score will be embodied by an elderly choir. Today you’re seeing a prototype, so your feedback is part of the work.