By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Peter Attia MD268 ‒ Genetics: testing, therapy, editing, association with disease risk, autism, and more
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
CHAPTERS
- 0:00 – 3:25
Wendy Chung's Path as a Physician-Scientist
Peter Attia welcomes Wendy Chung, who is moving to Boston Children's Hospital as chair of pediatrics. She describes her roughly 20/80 clinical-to-research split, her combined MD-PhD training, and how early NIH summers working on PKU shaped her approach.
- •Moving to Boston Children's Hospital and Harvard Medical School
- •About 20% clinical and 80% research, with the work starting and ending with the patient
- •Combined MD-PhD between Cornell and Rockefeller
- •Early NIH summers working on phenylketonuria sparked her interest
- •Her MD-PhD start coincided with the announcement of the Human Genome Project
- 3:25 – 15:10
PKU: Clinical Features, Screening, and Treatment
Chung explains phenylketonuria as a recessive condition treated with a low-phenylalanine diet, and how newborn screening evolved from bacterial assays to tandem mass spectrometry. They discuss why phenotype and genotype are both needed, the aspartame warning, pregnancy risks, and ethnic distribution.
- •Untreated PKU can cause intellectual disability, but early diet restriction prevents it
- •Newborn screening moved from bacterial agar assays to tandem mass spectrometry, with DNA as a second data stream
- •Phenylalanine levels fluctuate, so DNA sequence and blood levels together improve accuracy
- •The diet is a lifelong titration, and pregnant women with PKU need particular care
- •More frequent in Ireland, with roughly 1 in 5,000 versus about 1 in 10,000 in the US
- 15:10 – 17:58
Gene Therapy for PKU and Metabolic Disorders
Asked whether PKU might be a gene therapy target, Chung explains that inborn errors of metabolism are considered easier targets because many genes act in the liver. Carriers are indistinguishable from non-carriers, so a therapy doesn't need to reach 100% correction.
- •Liver-expressed metabolic genes are comparatively easy gene therapy targets
- •Recessive loss-of-function conditions only need some of the missing protein restored
- •Carriers with one working copy tend to be fine, which sets the bar for therapy
- •Chung wouldn't be surprised to see PKU gene therapies tried within our lifetimes
- 17:58 – 28:53
From the Double Helix to the Human Genome Project
Attia and Chung look back at how genetics was done before genome sequencing: manual PCR, radioactive gels, linkage maps, and positional cloning. They cover the structure of DNA, early gene discovery through proteins and enzymes, and the data-sharing ethos of the Human Genome Project.
- •Early PCR meant a graduate student moving samples between temperature baths by hand
- •Genome work was split across centers by chromosome, with strong data sharing
- •Genes were first identified through proteins and enzyme activity, as with sickle cell
- •Positional cloning, championed by Rudy Leibel, found genes by location alone
- •Her first gene cloned took eight years and her last took eight hours
- 28:53 – 35:09
Surprises from the Human Genome: Gene Count, Coding DNA, and Dark Matter
Chung discusses how the genome came in at about 20,000 genes versus the expected 100,000, with complexity coming from isoforms. They cover 99.9% human similarity, the roughly 1.5% coding fraction, and how contigs and repetitive regions left parts of the genome hard to read.
- •Estimates of up to 100,000 genes dropped to about 20,000
- •One gene can yield a dozen or more products through isoforms
- •Humans are 99.9% identical, yet one base change can mean life or death
- •Only about 1.5% of DNA codes for protein; the rest includes regulation and unknown function
- •Repetitive regions are a cryptic dark matter, only recently read telomere to telomere
- 35:09 – 44:29
Types of Genetic Tests: Consumer, Targeted, Exome, and Genome
Chung explains that genetic tests serve different purposes: ancestry and recreational tests versus medically targeted tests, exomes, and genomes. She also describes how a whole genome sequence is generated and interpreted, including the limits of current knowledge of gene-disease links.
- •23andMe and Ancestry are not medically targeted, so a clear result is not an all-clear
- •Medical uses include carrier screening for conditions such as Tay-Sachs and cystic fibrosis, and cancer risk testing
- •Chung distinguishes genetic from genomic, and exome from genome
- •About 7,000 of 20,000 genes have assigned disease associations
- •Short-read sequencing is assembled against a reference and interpretation is complex
- 44:29 – 52:31
Mosaicism, Germline Mutations, and Parental Age
Responding to Attia's own mosaicism finding, Chung explains somatic and germline mosaicism and how mutations accumulate over life. They compare how age affects eggs, with chromosomal aneuploidy, and sperm, with a linear rise in de novo mutations.
- •Genomes change over life, and cancer is largely driven by acquired changes
- •Germline mosaics can pass mutations on even when few cells carry them
- •De novo mutations rise linearly with paternal age
- •Maternal age drives chromosomal errors such as Down syndrome, with an inflection in the mid-thirties
- •Female meiosis begins in fetal life, so its proteins must work for decades
- 52:31 – 1:04:12
Exomes, Read Depth, and Breast Cancer Genetic Testing
They compare exome and genome sequencing, the limits of short reads, and how read depth corrects errors. The discussion moves to breast cancer testing options, the split among monogenic, moderate-risk, and polygenic contributions, and why liquid biopsy is not yet ready for screening.
- •Exomes were a cost shortcut, and the shift is toward genomes as costs fall
- •Read depth of about 30X filters out sequencing errors, and up to 1,000X is used for somatic mutations
- •Testing can be a few known spots, about 10 genes, or a panel of about 50
- •About 5% of cancers come from highly penetrant single genes, with another ~5% from moderate-risk genes plus polygenic risk
- •Cheek or blood samples capture the germline, while liquid biopsy is not ready for population screening
- 1:04:12 – 1:09:42
What Direct-to-Consumer Tests Actually Measure
Chung explains that consumer companies mostly read selected SNPs for ancestry and recreational traits rather than medically actionable information. Attia raises miscalled APOE results, and Chung advises confirming results before irreversible decisions.
- •Consumer tests read selected single nucleotide polymorphisms rather than the whole genome
- •Most variants examined are innocent bystanders useful for ancestry
- •Some SNPs can be medically impactful, since location matters
- •Lab errors such as plate flips can cause sample switches
- •Chung says to get a second opinion before major decisions such as a mastectomy
- 1:09:42 – 1:21:58
The GUARDIAN Study and Genomic Newborn Screening
Chung describes the GUARDIAN study, which sequences newborn dried blood spots for treatable conditions, building on her SMA and Duchenne screening work. They discuss uptake, diversity, cases caught, funding, and cost per baby.
- •SMA screening led to universal newborn screening and approved treatments including a one-time gene therapy
- •GUARDIAN stands for Genomic Uniform screening Against Rare diseases in all Newborns and screens only consented, actionable genes
- •About 74% of parents approached opted in, with over 3,000 babies screened
- •It found cases missed by traditional screening, such as a baby at risk for SCID
- •Cost is about $1,000 per baby with a goal of $100, and NIH funding is too small for a study of this scale
- 1:21:58 – 1:26:28
Which Diseases Have Gene Therapies and Which Need Them Most
Attia asks how many single-gene conditions have approved gene therapies, and Chung says only a handful, with SMA the standout. She notes that treatments like zinc for Wilson disease show gene therapy isn't always needed, and names Tay-Sachs, fragile X, and liver metabolic disorders as priority targets.
- •Only a handful of gene therapies are approved, including for SMA, Duchenne, and hemophilia
- •Wilson disease is treated cheaply with zinc once diagnosed early
- •Tay-Sachs fits the criteria of high prevalence, devastating course, and no treatment
- •Fragile X needs a different technical strategy
- •Liver-affecting inborn errors such as maple syrup urine disease and propionic acidemia remain undertreated
- 1:26:28 – 1:38:50
How Gene Therapy Works: Vectors, Gelsinger, and Sickle Cell
Attia and Chung discuss the death of Jesse Gelsinger, an adenovirus vector, and the immune responses that make dosing hard. They then cover why viruses are used, why stem cells must be targeted, and why gene addition doesn't suit sickle cell disease.
- •Gelsinger's treatment targeted the liver with an adenovirus vector that triggered an overwhelming immune response
- •Gene therapy often gives one chance, so dosing can be neither too high nor too low
- •Newborns with SMA have not met common viruses, which lowers the risk of immune reactions
- •Non-viral delivery such as lipid nanoparticles is being explored
- •Sickle cell needs stem cell editing or raising fetal hemoglobin, not gene addition
- 1:38:50 – 1:47:38
Gene Editing, CRISPR, and Strategies for Tay-Sachs and Fragile X
Chung explains gene editing from base editors to CRISPR-Cas9 and prime editing, including off-target effects from double-strand breaks. She then contrasts gene addition for Tay-Sachs with the repeat-shrinking editing needed for fragile X, and the challenges of delivery to the brain and timing.
- •Single-nucleotide changes are easiest to edit, and heterogeneous mutations are harder
- •CRISPR-Cas9 cuts both strands, with off-target effects as its Achilles' heel
- •Single-strand nicking approaches such as prime editing offer higher fidelity
- •Tay-Sachs could be treated by adding the enzyme back, but delivery throughout the brain is hard
- •Fragile X requires shrinking a trinucleotide repeat, and the treatable window may require very early intervention
- 1:47:38 – 1:55:55
The Genetics of Obesity and Epigenetics
The conversation turns to complex disease, starting with obesity, which has roughly 50% heritability but is strongly shaped by environment. Chung discusses leptin pathway mutations as rare but biologically informative, and cautions that epigenetic explanations remain largely conjecture.
- •Obesity heritability is around 0.5, and type 2 diabetes is more heritable
- •Pima Native American groups in different environments show environment's role
- •Leptin and its receptor are rare causes but reveal how body weight is regulated
- •Final common pathways may allow one-time genetic versions of drugs like statins
- •Epigenetics is hard to study across tissues, and the agouti mouse folate studies are suggestive but not conclusive
- 1:55:55 – 2:10:24
The Genetics of Autism
Chung covers autism's high heritability of about 0.8, de novo variants, and the hundreds of genes involved, many active in fetal brain development. They discuss heterogeneity, prevalence of about 2%, comorbidities, the gap in adult research, and transmission risk.
- •Heritability is estimated at about 0.8 to 0.9, but de novo variants are not captured in that measure
- •At least 100 genes are confidently linked, with perhaps 1,000 in total, and none accounts for more than 1% of cases
- •Genes function at the synapse or regulate gene networks, and some overlap with intellectual disability and epilepsy
- •Prevalence is about 2%, influenced by changing criteria, recognition, and access to resources
- •Autism is more heritable than bipolar disorder, schizophrenia, or depression, and little is known about adults
- 2:10:24 – 2:14:03
Cardiovascular Genetics and PCSK9-Style Therapies
Attia raises FH and Lp(a), and Chung emphasizes final common pathways and genetic cardiomyopathies, where early animal data look promising but the heart is a risky organ to treat. They discuss gene editing that mimics drugs, such as PCSK9 silencing, which would likely begin in extreme-risk patients.
- •Many genetic contributors to coronary disease may converge on shared biology, so therapies target the pathway
- •Genetic cardiomyopathies affect about 1 in 500 people
- •Early animal data on cardiomyopathy therapies look promising, but Chung is wary of gene therapy in the heart
- •Attia suggests editing PCSK9 could mimic a drug and cover heterogeneous causes
- •First trials would start with extreme-risk patients
- 2:14:03 – 2:21:37
Cost and Economics of Gene Therapy
Chung notes that current gene therapies cost one to three million dollars, which is unaffordable at population scale. Attia contrasts this with end-of-life chemotherapy costs, and they discuss fixed versus marginal costs, the field's infancy, and the potential for step-change cost reductions like those in sequencing.
- •Gene therapies currently cost roughly $1–3 million
- •Attia compares this with a million dollars of end-of-life chemotherapy for a year of extra life
- •About 10% of the US population has a rare genetic condition, raising scalability questions
- •High fixed costs with lower marginal costs suggest some savings are possible
- •Chung expects step-function breakthroughs, drawing on the COVID mRNA vaccine experience
- 2:21:37 – 2:35:33
Ethics of Gene Editing and the Future of Clinical Genetics
Prompted by Attia, Chung describes the scientific consensus against germline editing and enhancement and recounts the CCR5 case in China. They discuss embryo selection as an alternative, and Chung's outlook for 2040, in which diagnosis is trivial but therapy timing is uncertain.
- •Consensus is to treat somatic cells and not make transmissible germline changes
- •Enhancement is treated as a line not to cross, partly because of uncertain off-target effects
- •The CCR5 case was seen as enhancement where effective alternatives existed
- •Embryo selection can avoid editing, though some ethicists consider edge cases
- •By 2040 diagnosis may be trivial through cheaper sequencing and AI, but therapy rollout is uncertain
