By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Peter Attia MD290 ‒ Liquid biopsies for early cancer detection, the role of epigenetics in aging, and the more
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
CHAPTERS
- 0:02 – 7:05
From Engineering PhD to Medicine: Alex Aravanis's Background
Peter Attia and Alex Aravanis recall starting med school together as engineers. Alex describes his electrical engineering PhD in a neuroscience lab, where he studied synaptic vesicles, and the engineering principles he took from it.
- •Alex came in as an electrical engineer and applied signal processing and imaging to biology
- •His PhD work addressed how a brain synapse with few vesicles transmits so much information, the 'kiss and run' idea
- •Engineering lesson: if you can't build it, you don't understand it
- •The lab emphasized first-principles thinking and experiments that always teach you something
- 7:05 – 15:56
Joining Illumina and How Genome Sequencing Works
Alex explains why he was recruited to Illumina in 2013 and gives a primer on DNA, chromosomes and what it means to sequence a genome. He covers the public and Celera efforts of the Human Genome Project.
- •Illumina makes most of the world's DNA sequencing technology; Alex was hired to develop clinical applications
- •The genome is three billion bases packaged into chromosomes via nucleosomes and chromatin
- •The public effort used larger cut sections, while Celera's shotgun approach used small fragments reassembled computationally
- •First drafts of the human genome appeared in late 2000 and early 2001
- 15:56 – 24:41
Next-Generation Sequencing and the Collapse in Cost
The conversation turns to how the cost of a genome fell from hundreds of millions of dollars to a couple hundred. Alex credits miniaturization and sequencing-by-synthesis chemistry, and they compare this with Moore's law.
- •Cost fell by four or five logs between the genome project and 2012
- •Miniaturized, massively parallel reactions on a glass slide enabled billions of simultaneous reads
- •Sequencing by synthesis reads one base at a time across the whole 'lawn' of fragments
- •A genome now costs a couple hundred dollars, a faster decline than Moore's law
- •Next-gen sequencing now generates about 99% of sequencing data
- 24:41 – 35:32
From Tumor Sequencing to Liquid Biopsy and Cell-Free DNA
Alex explains how tumor panel sequencing led to the idea of sequencing tumor DNA from blood. He describes cell-free DNA, its roughly 160 base pair length protected by nucleosomes, and the three applications of liquid biopsy.
- •Tumor sequencing panels guide targeted therapy choices
- •Cell-free DNA was identified in the 1940s and found elevated in cancer patients in the 1970s
- •Dying tumor cells shed DNA into the blood, and fragments are about 160 bases long because nucleosomes protect them
- •Applications are screening, therapy selection and minimal residual disease monitoring
- 35:32 – 51:05
Prenatal Testing and the Accidental Discovery of Cancer in Cell-Free DNA
Alex describes how non-invasive prenatal testing works and how it incidentally detected cancer in pregnant women. Peter and Alex discuss why a small percentage shift is a detectable signal and where cell-free DNA comes from.
- •Prenatal testing counts fragments per chromosome bucket, and a trisomy raises chromosome 21 by only five to ten percent
- •Some women showed chaotic changes across many chromosomes and turned out to have cancers, reported in a 2015 JAMA paper
- •The findings were an inadvertent proof of concept, though not a screening test
- •Cell-free DNA comes from normal cell death and turnover, with only a few thousand cells' worth in a 10 mL tube
- 51:05 – 59:19
Founding GRAIL and the Specificity Challenge
Alex explains why no broad blood test for cancer existed and why specificity was the obstacle. He describes the reasoning that led to the spin-out of GRAIL in 2016 and its research plan.
- •Earlier biomarkers like PSA were sensitive but had false positive rates of 5 to 10 percent
- •A screening test needs roughly a 1% or lower false positive rate
- •Unknowns included how much DNA early tumors shed and the heterogeneity of tumor mutations, which pointed to deep sequencing and machine learning
- •GRAIL was spun out in early 2016 and the plan was to test every possible cell-free DNA feature head-to-head
- 59:19 – 1:05:26
The CCGA Study and Testing Every cfDNA Feature
Alex and Peter go through the features examined in the Circulating Cell-free Genome Atlas study: mutations, chromosomal changes, fragment size, cell-free RNA and methylation. They also cover how the study was run.
- •CCGA enrolled about 15,000 people, roughly 60% with cancer
- •The research phase took close to three years and cost hundreds of millions of dollars
- •Cancers often show chromosome doubling and rearrangement, which can be read from DNA per chromosome
- •Cell-free RNA is present, protected by proteins or vesicles such as exosomes
- •A blinded training set and test set design was used
- 1:05:26 – 1:17:22
DNA Methylation Explained and Why It Won
Alex explains methylation as an epigenetic 'software' layer and how bisulfite sequencing reads it. He then describes why methylation outperformed other features, and why more blood would not help screening.
- •Methylation is a methyl group added to cytosine at CpG sites, about 28 million positions, and controls gene expression
- •Bisulfite treatment converts unmethylated cytosines so methylation status can be inferred from sequence
- •Background biology, not blood volume, limited screening sensitivity
- •Methylation was the 'dark horse' but was most sensitive, highly specific and could predict tissue of origin
- •Other features added nothing because they were correlated with the methylation signal, shown through a Ford logo analogy
- 1:17:22 – 1:33:27
Sensitivity, Specificity and Positive Predictive Value in Screening
Peter uses the airport metal detector analogy and Alex explains the algorithm threshold that trades sensitivity against specificity. They discuss the Pathfinder study's roughly 40% positive predictive value compared with mammography and what that means for patients and physicians.
- •A threshold in the algorithm sets the trade-off, and methylation separated cancer from non-cancer unusually well
- •In Pathfinder, positive predictive value was around 40%, with prevalence under 1%
- •About half of cancers were detected, and the test roughly doubled cancers found over standard screening
- •Mammography's positive predictive value is about 4.5%
- •Peter says physicians should explain false positives and follow-up to patients beforehand
- 1:33:27 – 1:41:06
Which Cancers the Galleri Test Detects and Why
After Alex explains his own career path after GRAIL, the discussion turns to performance across cancer types. Alex frames tumor DNA detection as a functional assay tied to dangerous, fast-growing tumors, and Peter cites differences in breast cancer subtypes.
- •Alex's trajectory: Illumina, GRAIL co-founder, CTO at Illumina after it acquired GRAIL, then a new company as CEO
- •Sensitivity is lower for early prostate and hormone receptor positive breast cancers
- •Screening involves both over-diagnosis and under-diagnosis, and 70% of cancer deaths come from unscreened cancers
- •Kaplan-Meier curves show undetected cancers have much better outcomes than detected ones
- •Peter cites roughly 20% versus 75% sensitivity for early hormone-positive versus triple-negative breast cancer
- 1:41:06 – 1:54:41
Does Early Detection Improve Survival? The NHS Trial
Peter states his view that early detection improves outcomes and raises the skeptic's objection. Alex argues the stage-survival evidence is strong and describes the NHS pilot using reduction in stage four cancers as an endpoint, followed by a discussion of all-cause mortality.
- •Five-year survival is 70 to 80% for localized solid tumors versus under 20% for metastatic disease
- •Alex says it is hard to justify withholding treatment of cancers found by this test
- •The NHS pilot of about 140,000 people uses stage four reduction as a surrogate endpoint
- •Peter says screening trials are rarely powered for all-cause mortality, but a multi-cancer study offers more chances to see an effect
- 1:54:41 – 2:01:58
Methylation as a Hallmark of Aging and Epigenetic Reprogramming
Peter introduces his involvement in Alex's new company and asks where methylation fits in aging. Alex frames the epigenome as a readable, editable layer and poses the question of whether flipping methylation states back would restore youthful cell function.
- •Methylation sites are generally lost with age, and patterns differ with smoking, exercise and obesity
- •Alex mentions a talk suggesting methylation may be the 'hallmark of hallmarks' of aging
- •Hypothesis: restoring a young methylation state could make the cell behave as it did 25 years earlier
- •The epigenome can be read out and increasingly edited directly, giving an engineering approach
- 2:01:58 – 2:07:13
The Methylome Atlas and Tissue of Origin
Alex explains how the cancer site of origin prediction works and how the methylome atlas was built with a collaborator in Jerusalem. He notes the atlas also suggests comparing good and bad cell states to identify sites to fix.
- •Galleri predicts the likely cancer location, about 90% accurate with top two choices
- •Methylation patterns are copied when cells divide, preserving cell identity
- •The atlas was built from surgical remnants of healthy tissue with isolated cell types
- •Only the beta islet cell has an unmethylated insulin promoter
- •Comparing healthy and diseased methylation could reveal sites to fix
- 2:07:13 – 2:21:30
Yamanaka Factors, Partial Reprogramming and the Future of Rejuvenation
Alex explains Yamanaka factors and induced pluripotent stem cells, then partial reprogramming as a route to rejuvenation. They end with his optimism about organ-specific rejuvenation therapies within a decade.
- •Four transcription factors can revert differentiated cells to a stem cell-like state, but cell-type features are lost
- •Prolonged exposure risks making cells cancer-like and uncontrolled
- •Partial reprogramming in mice showed improved function in eyesight and cognition, though translation to humans is unknown
- •Alex expects rejuvenation therapies for tissues such as joints, retina and immune cells within ten years, not whole-body rejuvenation
