Summary #1214

Machine learning reveals recipe for building artificial proteins
  • Using machine learning models to analyze protein information from genome databases, researchers learn how to design artificial proteins with specific functions.
  • The models analyzed data about chorismate mutase proteins, allowing researchers to learn their design and encode it into synthetic genes.
  • The genes were cloned into bacteria which built the proteins—they had the same functions as natural proteins, and they performed them even better than natural.
  • Researchers don’t yet understand how their machine learning models uncovered the design rules of proteins, but it enables them to create any type of proteins they want.
  • Study authors are commercializing the technology, and hope that it will revolutionize fields like energy, environment (car


  • Science Science (69) How to precisely edit mitochondrial DNA
  • Psychology Psychology (76) Narcissists don’t learn from their mistakes because they don’t think they make any, study shows
  • Cosmos Cosmos (34) Ringing of global atmosphere confirmed in UH research
  • History History (18) Study settles the score on whether the modern world is less violent
  • Society Society (80) Dutch national broadcaster saw ad revenue rise when it stopped tracking users. It's meant to work like that, right?
  • Medicine Medicine (77) COVID-19 vaccines by Oxford, CanSino and Pfizer all trigger immune responses
  • Technology Technology (87) Machine learning reveals recipe for building artificial proteins
  • Nature Nature (57) Why are plants green?



Fame 🙌 - Articles for science lovers shortened to five bullet points. | Product Hunt Embed

We were featured on Hacker News, O'REILLY® Ideas, and Boing Boing.