#133 – Manolis Kellis: Biology of Disease

Episode Summary

Executive Summary: Manolis Kellis argues that human disease is becoming a problem of systems biology, not single genes: common, rare, and somatic variants converge on tissues, cell types, pathways, and regulatory circuits. He explains how human genetics, epigenomics, single-cell sequencing, and CRISPR-based perturbation are revealing unexpected mechanisms and enabling a faster path from variant to target to therapy.

Main Topics: Genetics as the new engine of basic biology (Priority: 5/5): Kellis explains that human genetics now drives discovery, reversing the older model where model organisms led and humans followed. Genetic variation in large populations provides natural perturbation experiments that reveal causal biology. Layered disease mechanism: variant → tissue → cell type → pathway (Priority: 5/5): The conversation lays out a pipeline for understanding disease by moving from genetic association to epigenomic enrichment, then to cell-type specificity, gene targets, and ultimately intervention points. Why noncoding variation matters (Priority: 5/5): Most disease-associated variants lie outside protein-coding regions, so the key challenge is identifying which genes and regulatory elements they affect through long-range chromatin interactions and enhancer wiring. Case study: obesity and the FTO locus (Priority: 5/5): Kellis details how the strongest obesity association near FTO was shown to act through a distant regulatory circuit involving a causal SNP, a transcriptional regulator, and IRX3/IRX5, altering fat-cell thermogenesis. Disease convergence and systems medicine (Priority: 4/5): Different genetic lesions often converge on the same few biological processes, making diseases more understandable as network problems. This supports a shift from single-gene medicine to pathway- and system-level intervention. Single-cell and multi-omic technologies (Priority: 4/5): He describes the technologies enabling the field: massively parallel reporter assays, single-cell RNA-seq, single-cell epigenomics, CRISPR perturbations, and computational integration across millions of cells. Therapeutic implications and future medicine (Priority: 4/5): The discussion ends with optimism that these tools will shorten the path from discovery to treatment, enabling cell-type-specific therapeutics, better prognostics, and eventually major disease alleviation.

Key Arguments: Human genetics now uncovers biological mechanisms directly in humans, often more powerfully than model organisms. Disease understanding requires breaking down the causal chain into tissues, cell types, regulatory elements, genes, and pathways. Most disease variants are noncoding, so locus-to-gene mapping depends on 3D genome architecture and regulatory circuitry. Strong genetic effects are not always the best therapeutic targets; modest-effect nodes in key pathways can be more druggable. Disease biology shows convergence: many distinct variants funnel into a limited set of core processes. Single-cell multi-omics reveals cell-type-specific disease programs that bulk tissue averages hide. CRISPR and pooled assays make it possible to test thousands to millions of hypotheses in parallel. The future of medicine is systems-level and personalized, integrating genotype, cell state, and electronic health records.

Data Points: Genetic perturbations per person: ~6 million - Kellis says each human carries roughly six million inherited variants that can be viewed as natural experiments. Human population size referenced: ~7 billion people - Used to illustrate the scale of genotype-phenotype data available for study. Coding fraction of genome: ~1.5% - He notes that only a small fraction of the human genome codes for proteins, leaving most disease variants in noncoding DNA. Disease variants outside protein-coding regions: 93% - Most disease-associated variants are said to fall outside genes/protein-coding sequence. Genes in the human genome: ~20,000 - Used when discussing the challenge of linking variants to the genes they regulate. Human genome length: 3.2 billion nucleotides - Referenced multiple times when describing variant counts and the scale of genome editing. Strong Alzheimer's heritability estimate: ~79% - Kellis states Alzheimer's has a large genetic component, though environment still matters. Major U.S. mortality from heart disease: 650,000 deaths/year - Used in a discussion of disease importance and public-health burden. Major U.S. mortality from cancer: 600,000 deaths/year - Listed as the second leading killer in the U.S. in the conversation. Accident deaths: 167,000 - Referenced as number three among causes of death, despite lower media visibility. Lower respiratory disease deaths: 160,000 - Included in the mortality ranking discussed as a measure of disease importance. Alzheimer's deaths: 120,000 - Used to show Alzheimer's' major health burden. Stroke/aneurysm deaths: 147,000 - Mentioned in the same public-health ranking. Diabetes/metabolic disorder deaths: 85,000 - Part of the comparative mortality list. Flu deaths: 60,000 - Cited among annual U.S. deaths. Suicide deaths: 50,000 - Included as an example of a high-impact condition affecting quality of life and mortality. Alzheimer's-associated loci: 27+ - Kellis says more than 27 loci are associated with Alzheimer's at the end-to-end genetic level. Epigenomic tissue atlas size: 127 tissues - The Roadmap Epigenomics consortium map used to connect variants to tissue-specific regulation. Next-generation atlas size: 833 tissues - Epimap, the expanded tissue resource, is described as the next generation of the atlas. Disorders surveyed with the expanded atlas: 540 disorders - He says enrichments were found across hundreds of disorders using the 833-tissue map. Single-cell dataset generated in one year: 10 million cells - Kellis claims his team generated ten million human brain cells across multiple disorders in a year. Brain regions sampled: 12+ - Referenced in describing the scale of the single-cell brain dataset. Individuals sampled in brain work: 1,500 - He states the dataset spans brains from about 1,500 individuals. Common obesity-associated variants near FTO: 89 - In the obesity case study, 89 common variants were considered in the locus. Distance from FTO locus to target genes: 1.2 million nucleotides - The obesity-associated variant was said to regulate IRX5 at this distance. Distance to IRX3: ~600,000 nucleotides - The same variant also regulates IRX3 from a long genomic distance. Cool bed temperature claim: 55°F - Mentioned in the sponsorship read for the 8 Sleep mattress.

Pivotal Quotes: "Human genetics has been so transformed in the last decade or two that human genetics is now actually driving the basic biology." — Manolis Kellis: Explaining the reversal of the traditional model-organism-first approach. "We are now at the level of systems medicine." — Manolis Kellis: Summarizing the move from single-gene medicine to network- and pathway-level intervention. "The genetic variant sits in the first intron of the FTO gene, but it controls two genes, IRX3 and IRX5, that are sitting one point two million nucleotides away." — Manolis Kellis: Describing the obesity locus that showed long-range regulatory control rather than a simple gene-level effect.

Implications: Listeners should expect biology and medicine to become more predictive, personalized, and network-based. For pharma and research, the key shift is from single genes to cell-type-specific regulatory circuits, enabling faster target discovery and more precise therapies.