Berkeley, California, United States of America
Ashley Zehnder
2018
Private company
Seed
4.1
Boom Capital Ventures, Pacific8, True Ventures, The Longevity Fund, Beagle Ventures
machine learning gene targets from animals
The era of inexpensive DNA sequencing offers an unparalleled opportunity to understand the causes and potential cures for human diseases, but studying humans alone is not enough. Translating findings from nearly two decades’ worth of human genome-wide association studies (GWAS) into therapies has been relatively unsuccessful because most diseases are caused by hundreds of mutations with small effects, while mutations with large effects in the key disease genes occur too rarely in humans to be detected. Identifying genes with large effects will lead to many disease cures, but they are hard to find with human data alone. Human studies of rare mutations have identified drug targets with large disease impacts, such as PCSK9 for cholesterol levels and have led to successful drugs. However, the scarcity of these mutations makes this strategy untenable for most diseases. A novel and more widely-applicable way to identify genes with large effects is to examine animals with medically-relevant traits.
The vast majority of human genes are shared with other mammals (92% shared with rodents) and perform very similar functions. Differences in the expression of these shared genes protect other mammals from diseases including muscle and bone loss, and can even reverse Alzheimer’s-like pathologies. Unlike humans, these animals undergo large physiological changes, and the shared genes driving these changes have larger and more easily detectable effects. At Fauna Bio, we identify genes in animals that are responsible for disease protection and that have strong connections to related human genes. This unique strategy allows us to pinpoint genetic targets with large effects on human disease.
Hibernating mammals are an excellent system for identifying drug targets because they can transiently protect themselves from problems similar to critically important human diseases, allowing us to identify key protective genes that are turned on or off at medically-relevant times. They are naturally protected from numerous physiological insults that would be lethal to a non-hibernator:
Our platform finds gene networks that turn on and off during disease protection time points in hibernators and verifies the translatability of potential targets using human genomics data and cross-species conservation.
We utilize computational drug design tools and deep learning-based approaches to predict compounds that will hit our targets.