Therapeutic Target Identification

therapeutic target identification schemeA critical challenge in target validation and drug discovery is the development of preclinical assays that predict whether a therapeutic will ultimately succeed in the clinic. In order for a cellular assay to predict clinical response, it must read out a biological function that, if perturbed by a drug, will yield benefit to the patient; i.e. disease-calibrated.  For most diseases, preclinical assays prove poorly predictive of clinical response, resulting in expensive failures in late-stage clinical trials. 

Human genetics offers a method to forge this critical connection between human disease phenotypes in vivo and preclinical assays in vitro — by assessing genetic variants both for association to disease in populations, and to functions in the laboratory, it should be possible to triangulate on disease relevant functions of cells. Genome sequencing of large cohorts now makes it possible to mine human genome sequence variation for “experiments of nature” that perturb the functions of a wide range of genes.  In some cases, these “experiments of nature” can be used to infer a dose-response curve of gene function that indicates how enhancement or suppression of the encoded protein's activity raises or lowers disease risk. 

Our goal is to identify genes which confer increased insulin sensitivity when inactivated by nature through loss-of-function mutation. Since insulin sensitivity is clinically challenging to quantify, we narrow the search space by utilizing genetic screens in insulin sensitive cell types (adipocytes, hepatocytes and myocytes) to find genes that enhance cellular insulin sensitivity when ablated. Prior to large scale screening these assays are disease-calibrated by tuning their ability to discriminate known mutations that increase/decrease insulin sensitivity in humans. The genes which emerge from these screens are then interrogated in population based exome sequencing datasets for loss-of-function genetic variation and correlated with the phenotypes of the people who carry them via electronic health record analysis and recall-by-genotype clinical investigation.

For more context about this research aim see Jiao et. al., Molecular Metabolism 2019.

Individually Tailored Therapy

individually tailored therapy scheme
Missense variants that alter protein function are a major cause of inherited disease with only a minority of disease mutations caused by stop codons, frameshifts, deletions and other severe changes to the encoded protein. Functional characterization of missense changes is necessary to provide molecular diagnosis, estimate risk, screen family members, and guide therapy. Traditionally, a variant was classified as pathogenic if it was identified in an individual with a Mendelian disease phenotype, segregated with disease in families, and demonstrated severe abnormalities in a laboratory assay of function. This process is too slow and resource intensive for clinical use, leading to many Variants of Uncertain Significance (VUS). With the advent of low cost clinical exome sequencing hundreds of VUS are being identified in genes previously implicated in a severe genetic disease. Every additional exome sequenced identifies ~200 novel protein-coding variants. 

We utilize modern synthetic biology techniques and massively parallel, disease-calibrated cellular assays to synthesize and test all possible missense variants at clinically important genes, i.e. saturation mutagenesis in mammalian cells. This prospective functional characterization enables protein coding VUS to be instantly interpreted for function and clinical effect. Our current efforts focus on utilizing saturation mutagenesis data to predict and guide treatment response in metabolic diseases.  

For more context about this research aim see Majithia et. al., Nature Genetics 2016.

Next-Generation Phenotyping

next generation phenotyping scheme
Prior to the invention of the EKG, it was difficult to distinguish separate etiologies of symptomatic chest pain with a rapid heartbeat. With the ability to continuously monitor the electrical activity of the heart, distinguishing arrhythmias from myocardial infarctions is now routine. Before the EKG became a diagnostic tool for heart disease it was necessary to map pathophysiologic states to electrical recordings. Continuous glucose monitoring sensors (CGMS) have exponentially increased the amount of glycemic data (from ~4-6 measurements a day to 288 measurements per day) that can be obtained from an individual. In doing so they provide the opportunity to obtain a deep and nuanced profile of diabetes and insulin resistance, but we lack an understanding of normal versus pathologic. Our efforts are focused on collecting CGMS profiles from normoglycemic individuals as well as those with varying types of pharmacologic (e.g. steroids, anti-retrovirals, anti-psychotics) and genetically (e.g. lipodystrophy) induced diabetic states. These reference profiles will be used to train machine learning algorithms to identify “glucotypes” in individuals currently classified as “type 2 diabetes” (T2D). Ascertained on larger cohorts, these glucotypes can be mapped to clinical outcomes providing meaningful T2D subtyping for improved prediction of secondary complications and tailoring of therapy.

For more context about this research aim see Majithia et. al., Journal of Diabetes Science and Technology 2018.