Research Topics

Transcriptional dissection of microglia in an Alzheimer’s Disease Mouse Model

Microglia heterogeneity has shown the spectrum of phenotypes displayed in response to a variety of insults and in several conditions. Our published work shows that an anti-TREM2 therapeutic antibody increases the proportion of proliferating and transitioning microglia and decreases the proportion of homeostatic microglia. The antibody decreases filamentous plaques and neurite dystrophy.

using CRISPR to understand Functional noncoding dNA

Since the massive effort of sequencing the entire human genome in 2001, we have found that ~98% of the human genome does not encode for proteins. In addition to investigating and cataloguing what the coding portion of DNA, the noncoding genome is equally, if not much less, explored. Studies have found more single base pair variants (SNPs) in noncoding DNA, compared to coding, in human diseases. These diseases includes Alzheimer’s, cancer, ALS, IBD and auto-immune diseases. Through many iterations of research in toolkits to edit DNA content (e.g. convert an adenine to a guanine), the CRISPR system was discovered. A beautiful history of CRISPR is narrated by Eric Lander in this article. Parallelized versions of CRISPR (CRISPR screens) were implemented in the early 2010s and have become a frontrunning technology to test many candidates simultaneously, rather than one-by-one. I use CRISPR screens to dig deep into what pieces of noncoding DNA control cancer cell growth.

Deep learning & CRISPR

Deep learning networks can help us understand which & why DNA is functional.

I combine CRISPR pooled screening technologies and computational analysis (machine & deep learning) to understand the function of noncoding DNA in human diseases. In the context of cancer, oncogenes cause unchecked proliferation and their regulation is a point of therapy. To better understand the regulatory architecture at the DNA loci, computational models predicting the functionality level gives insight into how biological systems use DNA to achieve their phenotype. This leads to better principles when designing mutagenic tiling screens through noncoding DNA. It may not be useful (or too time- & cost-inefficient) to investigate every letter of noncoding DNA, but a narrower, informed approach can give researchers a prior probability as to which segments of DNA contain valuable vs. useless information.