Kellis, Manolis
▶ Watch on YouTubeCore ideas
Kellis argues that human genetics has inverted the traditional order of biological discovery: instead of studying model organisms (mouse, fly, yeast) and extrapolating to humans, human genetic variation — every person carries roughly six million naturally occurring perturbations relative to the reference genome — now constitutes the largest "natural experiment" available, driving basic biology rather than following it. He distinguishes strong-effect "Mendelian" mutations (traceable through family pedigrees, but often so severe that evolution has purged most variation, making them individually rare) from weak-effect polygenic variants studied across large unrelated cohorts, and makes the case that weak-effect variants in more "druggable" genes are often better therapeutic targets than strong-effect ones, since a drug can safely induce a much larger perturbation than evolution has ever tolerated naturally. He also explains why the near-total absence of natural variation in a gene is itself informative: it signals the gene is so essential that mutations are eliminated before or shortly after conception (early embryonic lethality), rather than that the gene is unimportant.
A central methodological thread is why 93% of disease-associated genetic variants fall outside protein-coding regions, making "nearest gene" assumptions misleading — illustrated by his lab's widely discussed 2015 finding that the strongest known genetic association with obesity, long attributed to the FTO gene, actually acts on two entirely different genes (IRX3 and IRX5) sitting over a million nucleotides away, mediated by a regulatory element and the transcription factor ARID5B. He lays out a six-step general pipeline his lab uses to move from genetic association to mechanism: (1) identify the relevant tissue via epigenomic enrichment across large reference maps (833 tissues in his lab's current "EpiMap"), (2) pinpoint the single causal variant among many co-inherited ones in a linkage block, (3) find the regulatory "motif" it disrupts, (4) identify the regulator protein that normally binds there, (5) trace physical (chromatin folding/Hi-C), genetic (eQTL), activity-correlation, and CRISPR-perturbation links to determine the actual target gene(s), and (6) determine the target genes' downstream biological function. Applied to the FTO/obesity locus, this pipeline revealed that the causal mechanism runs through thermogenesis versus lipid storage in mesenchymal-stem-cell-derived fat progenitors — a mechanism his team validated by using CRISPR to flip a single nucleotide (out of 3.2 billion) in patient-derived fat cells and switch them between an obese and lean phenotype. He notes he personally carries the risk allele.
Kellis explains that this same enrichment-based approach revealed the immune system's outsized and often unexpected role across seemingly unrelated diseases: an immune "complement" pathway was independently discovered as central to both age-related macular degeneration and (via microglial synaptic pruning) schizophrenia, and microglia — the brain's resident immune cells — turned out to be the single cell type most enriched for Alzheimer's genetic risk, essentially absent from enrichment in neurons, oligodendrocytes, or astrocytes. He argues this kind of cross-disease convergence (many independent genetic "hits" scattered across the genome nonetheless converging on the same small set of biological pathways, such as calcium signaling in schizophrenia) is what makes the genomics-based approach tractable despite its enormous complexity, and that it necessitates dissolving traditional disease-siloed academic departments in favor of shared "circuitry" infrastructure used across many disorders simultaneously (his own lab works in parallel across cardiovascular disease, immune disorders, cancer, schizophrenia/autism/PTSD, Alzheimer's/neurodegeneration, and metabolic disease). The conversation surveys the technological toolkit enabling this: CRISPR (whose genuine innovation, Kellis stresses, is not the cutting or repair mechanism — both pre-existing, borrowed from bacterial antiviral immunity — but the ease of programming a guide RNA), massively parallel reporter assays that can test thousands of regulatory-element hypotheses simultaneously via DNA microarray synthesis, Hi-C chromatin-folding assays, and single-cell RNA/DNA sequencing (including a "bottle-based" combinatorial barcoding trick that uniquely labels a million cells using only 100 wells reused three times). He closes by describing frontier work in his lab — a 10-million-cell single-cell atlas across a dozen brain disorders, tracing Alzheimer's pathological spread region-by-region for early diagnosis, studying somatic (post-zygotic) mosaic mutations as a distinct third layer of genetic variation from common and rare inherited variants, and moving intervention design from single genes toward whole-pathway "systems medicine" — framing the ongoing convergence of genomics, single-cell technology, CRISPR-based perturbation, and machine learning as poised to fundamentally transform disease treatment within his professional lifetime.
Excerpts
On why the strongest genetic hit for obesity turned out not to involve the gene it sits in:
“This FTO locus, even though it is associated with obesity, does not implicate the FTO gene. The genetic variant sits in the first intron of the FTO gene, but it controls two genes, IRX3 and IRX5, that are sitting 1.2 million nucleotides away, several genes away... I was introduced at a conference as the guy who wrote the most depressing paper of 2015 — because the entire pharmaceutical industry was so comfortable that there was a single gene in that locus.”
On why convergence across scattered genetic "hits" makes the genomics approach tractable:
“You have one hit here and one hit there and one hit there and one hit there — completely different parts of the genome — but it turns out all of those hits are calcium signaling proteins... those people who are focusing on one [gene at a time] cannot possibly see that picture. You have to become a genomicist, you have to look at the omics, the holistic picture, to understand these enrichments.”
On why traditional single-gene, single-disease academic silos no longer make sense:
“The fundamental understanding of the circuitry of the human genome that you need to solve schizophrenia — that fundamental circuitry is hugely important to solve Alzheimer's, and that same circuitry is hugely important to solve metabolic disorders... for us to say, 'you guys in the immune building go solve the fundamental circuitry of everything, and you guys in the schizophrenia building go solve it separately' — is crazy.”