←
Fitzpatrick Laboratory3,309 posts
Fitzpatrick Laboratory avatar

Fitzpatrick Laboratory

@FitzpatrickLab

Assistant Professor @ZuckermanBrain @ColumbiaBiochem. Lab uses cryo-EM & biophysics to study the cellular, molecular, and structural basis of neurodegeneration.

◷ Joined January 2013▧ 98 Photos and videos
Laura Kida avatar
⟳ Fitzpatrick Laboratory retweeted
Laura Kida@LauraCKida·Nov 16

About 90% of RNA molecules in our cells are tRNAs. To date, we have mostly been blind to their abundance and regulation in single-cell sequencing. Today, I’m excited to share a new manuscript from my MD/PhD work biorxiv.org/content/10.64898…

Daniel Tawfik avatar
⟳ Fitzpatrick Laboratory retweeted
Daniel Tawfik@dantawfik·Nov 14

Twin studies say Alzheimer's is 60 to 80% heritable. Common DNA variants were only explaining about 5% of it. A new Nature study of 2.8 million people puts that figure at 16%. That gap has bothered geneticists for years. Some argued there was little genetic risk left to find beyond APOE, the best-known Alzheimer's gene. Others argued over 1,000 variants must be involved. This paper tests which camp was right. Here is what the researchers did. The Psychiatric Genomics Consortium Alzheimer's group pooled 30 cohorts. That included 109,479 people with Alzheimer's, 74,141 proxy cases, and about 2.6 million controls. Proxy cases are people who don't have the disease themselves but reported that a parent did. Each parent shares half your DNA, so you can use their diagnosis as a stand-in. This matters because Alzheimer's strikes late, and most biobank volunteers are still too young to have it. The study is a genome-wide association study (GWAS). It scans millions of common DNA spelling differences and asks which ones show up more often in people with the disease. Each hit marks a region of the genome, called a locus, that nudges risk up or down. Sample ancestry was 91% European, 5% African, 3% East Asian, and under 1% each admixed American and South Asian. What they found: • 127 risk loci in total, and 48 of them are new • A heritability estimate of 16% with a newer statistical method, versus 6% with the older standard method on the same data • A polygenic score (a single risk number built from thousands of variants) that explained an average of 12.8% of risk variation in European cohorts, and up to 16.9% in the best one • In people over 65, the top 10% of the score had a sevenfold higher Alzheimer's prevalence than the bottom 10%, and a threefold higher prevalence than the whole sample • Alzheimer's-linked genes that were more active in microglia (the brain's immune cells) in patients, with microglia the only cell type that held up after correction (P = 1.58 × 10^-6) • 17 of the new genes already interact with approved drugs for other conditions, though only five had been tested as Alzheimer's targets Why did the old estimate come in so low? Heritability is the share of differences in risk between people that comes from their DNA. The older standard way to estimate it from GWAS data is called linkage disequilibrium score regression, or LDSC. It works by checking how strongly each variant's signal tracks with its neighbors on the genome, and it assumes risk comes from thousands of variants, each with a tiny and similar effect. Alzheimer's doesn't look like that. One variant, APOE-ε4, explains about 6% on its own. Think of it like estimating a choir's volume by assuming every singer is equally quiet. A single soloist breaks the assumption, and the total comes out too low. The authors tested this directly. They simulated a trait with an Alzheimer's-like architecture in UK Biobank genotypes, with one large-effect variant plus 1,000 small ones. LDSC underestimated the true heritability. The newer method (SBayesRC), which allows for a mix of big and small effects, recovered it once the APOE region was handled separately. A second check points the same way. The polygenic score explained more variance (about 13%) than LDSC said was theoretically possible for all common variants combined. So what do the new genes tell us about biology? The paper's logic runs in three steps. First, they matched each locus to its most likely gene. They combined machine learning predictions, links between DNA variants and gene activity, and earlier exome studies. Second, they asked which brain cell types use those genes most. Microglia kept coming up. Genes turned up in microglia in patients were enriched for genetic risk. Genes turned down in certain neurons were too, though that neuron signal weakened to nominal after correcting for general neuronal downregulation. The authors say it may not be specific. Third, they grouped genes by pathway. Immune activation, lipid binding, amyloid breakdown, and cell adhesion returned, as in earlier studies. Five processes were new to GWAS, including glutathione metabolism (oxidative stress handling) and sphingolipid metabolism (fats that shape cell membranes). A few genes stood out as drug candidates. SYK and AXL help microglia respond to amyloid. QPCT makes the exact amyloid modification that donanemab, an approved antibody, targets. Eight of the 48 new loci couldn't be tied to any known pathway, which hints at biology nobody has described yet. Now the limits. The data are 91% European. APOE-ε4 effect sizes ranged from an odds ratio of 1.7 in African ancestry to 3.4 in East Asian ancestry. But even European cohorts varied a lot, so cohort differences may explain much of that spread. The polygenic score also performed worse outside European ancestry, so it can't yet be applied evenly across populations. The 16% figure assumes a 5% lifetime prevalence. Assuming 20%, which is closer to the rate in people over 84, it rises to roughly 25%. The number depends on that choice. Proxy cases make up 40% of all cases, and parent reports can be wrong. The authors report that proxy and clinical results agreed closely (genetic correlation 0.99), but diagnosis accuracy is a real caveat. These are associations. A locus points to a region, and the gene assignment is a prediction. The drug target list is a hypothesis list, and two of the five tested targets failed in earlier human or mouse work. Finally, the polygenic score isn't a diagnostic. It shifts group-level risk. Most new variants change odds by only a few percent each. Prediction at the individual level is still limited. What does this mean for you? If you have a family history, your risk is genuinely shaped by DNA, and more of it is common variation than people thought. But common variants are not destiny. Each one nudges risk slightly, and none replaces the lifestyle and vascular factors that also matter. For the field, the finding that microglia keep appearing is a push to look beyond amyloid removal, which has delivered limited symptom relief so far.

Post attachment
Gene Yeo’s Lab at UC San Diego avatar
⟳ Fitzpatrick Laboratory retweeted
Gene Yeo’s Lab at UC San Diego@yeo_lab·Nov 4

New in @NatStructMolBio! 🧬 We screened 730 RNA-binding proteins and found 281 that control circular RNAs, from how they’re made to how much protein they produce. Thanks to our collaborators at MSK, Columbia, and UCSD, and to my amazing lab, led by Tao Yu nature.com/articles/s41594-0…

Fitzpatrick Laboratory avatar
Fitzpatrick Laboratory@FitzpatrickLab·Oct 26

Protein fibers, a hallmark of Alzheimer’s, spotted making deliveries to sperm, via @EurekAlert! eurekalert.org/news-releases…

Molecular Cell avatar
⟳ Fitzpatrick Laboratory retweeted
Molecular Cell@MolecularCell·Oct 26

Online Now: Amyloid-mediated RNA uptake by sperm for embryonic delivery dlvr.​it/TW44Xs

Post attachment
Columbia University's Zuckerman Institute avatar
⟳ Fitzpatrick Laboratory retweeted
Columbia University's Zuckerman Institute@ZuckermanBrain·Oct 26

Amyloid fibers are best known as hallmarks of Alzheimer’s and other neurodegenerative diseases. But they recently turned up in a surprising place: in sperm, shuttling genetic material. The new research from the @FitzpatrickLab holds potential for gene-editing technologies. Learn more from @MolecularCell: zuckermaninstitute.columbia.… @ColumbiaMed