Samuel King

But when I started taking classes as an undergrad, I realized that biology is best broken down at the molecular level. I was particularly intrigued by the intersection of biology and design, and I became obsessed with understanding DNA, RNA, and proteins, which are the substrates of life that evolution has optimized for nearly 4 billion years. It became clear to me that information at that scale of biology quickly becomes too much for humans, and that machine learning is one of the best ways for us to understand biology to its fullest. It could take a human being a lifetime to understand one genome, but a machine learning model could learn features from millions of genomes within a matter of weeks. When I looked into bioengineering programs that combined all of my interests, Stanford’s stood out because it felt really comprehensive and modern, like the future of the field would be determined here. 

I’ve never had as many resources available to me as I do now as a second-year PhD student in the bioengineering department. I’m a member of Professor Brian Hie’s lab, the Laboratory of Evolutionary Design, and we have built machine learning models, which we call Evo 1 and Evo 2, that have been trained on almost all observed evolution. Evo 1 and Evo 2 are really great at predicting things about DNA and also designing DNA. For example, if you had a gene that had a cancerous mutation in it, Evo might be able to recognize that and classify it as undesirable. Evo has a specific model architecture called StripedHyena, which excels at long-range understanding of sequences and allows it to outperform existing models that have attempted this type of processing before. 

I use Evo to design entire genomes that are small enough to test in the lab. The design process is fully computational, and then we synthesize the DNA and test that the genomes are functional in real life. This is a rigorous application of Evo and pushes synthetic biology to its limits, which may benefit biotechnology a lot. It’s amazing that we’ve trained these machine learning models well enough that we can drive evolution beyond what has occurred in nature. Even though I’m working through this molecular language of biology, it feels like I’m still connected to what originally awed me about being outdoors and the natural world. We’re basically doing evolution inside a computer, which is really cool.

Something that I think is important about our work is that we talk about ethics a lot. We work closely with the Boussard Lab, a Stanford-led bioethics group, and they were very helpful in making sure that Evo was created in a safe way. One big decision we made for Evo 1 and Evo 2 was to exclude any eukaryotic virus genomes from the training data to prevent misuse of the model such as for designing infectious viral components as bioweapons. We want to help humans with our work, and Evo has applications across all aspects of biology, especially biomedical. 

A big part of being a scientist for me is figuring out ways to communicate the science that I do well so that people can understand it and be as excited about it as I am. I have also always loved art and graphic design, and I try to integrate both into the science that I do because I think scientific communication and scientific visualization are really important. 

Coming to Stanford, I was immediately struck by how unparalleled it felt in terms of its resources. It’s central to the future of science and is a hub where great scientists gather. I feel so lucky to be a part of what’s happening here because I’m getting to do biological design at the scale that I always dreamed about.

I’m not sure what the future holds yet but I am drawn to the idea of leading a research-based nonprofit or startup someday. For now, I’m really happy being at Stanford and using any free time I have to explore my surroundings, play music, or hang out with friends.