Current work

I’m currently developing a new Alzheimer’s drug by myself. Over the past year, I used AI to design several thousand new small molecule drugs. I taught myself synthetic chemistry, built a chem lab in my garage, and synthesized a hundred of these drugs. I’m now testing them in cell lines and mice for efficacy towards a new Alzheimer’s mechanism.

All this was done using my savings from graduate school. This might seem impossible. Drug developers spend tens of millions of dollars, if not significantly more, generating pre-clinical efficacy evidence for a new drug. However, I have a new approach that can accomplish this for 1/1000 the usual cost.

My approach is heavily inspired by strategies developed during the “golden age” of drug discovery - the 1960s. Paul Janssen, the most prolific drug developer ever, operated during that time. He and a small chemistry team developed 80 new drugs over the course of his career. His approach was very different from what people do today. It emphasized cheap, rapid synthesis of analogs of existing drugs, rapid animal testing, no target binding studies, and disease mechanism selection based on intuition, among other things. His results don’t lie. Taking Janssen’s approach and augmenting it with modern biotechnology and AI tools is how I’m doing it.

Stay tuned.

The first one-person unicorn will be a pharma company.

Past work

Compressed Perturb-seq

I invented a new genomics assay called “compressed Perturb-seq.” This is a ten-fold cheaper version of standard “Perturb-seq,” which involves modifying the DNA of cells using CRISPR then reading out changes in the mRNA expression levels of all ~20,000 expressed genes for each cell using single-cell RNA-sequencing.
Original publication: Yao et al. 2023, Nature Biotechnology.
Editorial highlights of the technology in Nature Biotechnology and Nature Genetics.
Blog post, Twitter thread, and video (credit to Julia Bauman) that summarize the paper in non-technical terms.

Mediated expression score regression

I invented a new statistical framework and inference method to quantify how much genetic effect on a given disease “flows” through gene expression measurements in large human populations.
Original publication: Yao et al. 2020, Nature Genetics.
Technical blog post, non-technical blog post, and Twitter thread that summarize the paper.

Other

I wrote this paper on cancer genomics when I was an undergrad: Yao et al. 2019, Scientific Reports.