Emil Uffelmann

I know that Galton called out for sharing data in 1901

Genomic analyses reveal new insights into Alzheimer's disease Alzheimer's disease (AD) is the most common cause of dementia, with global case numbers projected to reach 153 million in 2050. AD is highly heritable, with twin-based heritability estimates of 60-80%...

🚨 The preprint for our GWAS of Alzheimer’s disease (AD) is out

Preprint: medrxiv.org/content/10.1...

🧵

Proud that the third GWAS for Alzheimer's dementia from the PGC-ALZ working group was just posted online! Huge amounts of work, and what a great collaboration! Check out our exciting findings below 👇 @pgcgenetics.bsky.social #ctglab #alzheimer #dementia #GWAS

special thanks to Douglas Wightman (shared first author), my PIs @daniposthu.bsky.social & Ole Andreassen, and the whole Alzheimer's disease working group of the Psychiatric Genomics Consortium

This study was only possible thanks to a collaboration of >100 co-authors, biobanks, pharma partners, direct-to-consumer companies, and, most importantly, the nearly 3 million participants who shared their data. I am immensely grateful for their participation.

Genomic analyses reveal new insights into Alzheimer's disease Alzheimer's disease (AD) is the most common cause of dementia, with global case numbers projected to reach 153 million in 2050. AD is highly heritable, with twin-based heritability estimates of 60-80%...

The paper contains many more analyses and details. Check out the paper if interested.
Preprint: www.medrxiv.org/content/10.1...

While ~90% of our data were from Europeans, our multi-ancestry design and ancestry-specific summary statistics pave the way for more diverse future AD GWASs.

We also generated stratified GWAS results by sex, ancestry, and phenotype definition, and versions excluding the UK Biobank.
All summary statistics will be made publicly available upon acceptance to maximize reuse and transparency.

Our polygenic prediction models explained up to 17% and, on average 13% of variance in European cohorts. Given that SNP-heritability estimates are the theoretical upper limit for polygenic prediction, the LDSC estimates are clearly underestimates.

Interestingly, using LAVA, we estimate that the SNP-heritability of the APOE region alone is ~9%.

Using SBayesRC, which models a mixture of SNP effect sizes and can better account for large-effect variants, we estimated SNP-heritability at ~19% (vs. 6% from LDSC). Estimates were similar across African and East Asian ancestries.

AD GWASs have long been plagued by low SNP-heritabilities (~5%), far below the 60–80% twin-based estimates.

Sncg and Sst, two GABAergic neurons, have been previously shown to be vulnerable early on in the AD disease process, suggesting that AD-associated variants may influence gene expression in vulnerable neuronal subtypes, leading to neuronal cell death.

We found enrichment for:
Upregulated genes in microglia, and downregulated genes in three neuronal subtypes (Sncg, Sst, and L6 IT Car3).

We also went further:
Using differential expression between AD cases and controls, we tested whether up- or down-regulated genes in specific cell types were enriched for genetic signal.

Previous AD GWASs using gene expression in healthy controls linked genetic risk mainly to microglia. We replicate that finding: genes highly expressed in microglia show strong association with AD risk.

We identified 118 loci in a multi-ancestry GWAS and 9 more in a European-only GWAS (total = 127 loci).
Of these, 48 were novel, including 8 potential drug targets: QPCT, EGFR, KEAP1, SYK, AXL, RRM2B, CACNA1S, and IL23A.

Summary: We analyzed ~180K cases & 2.6M controls, identified 127 loci (48 new), improved heritability estimates (19% in Europeans) & PGS prediction (mean 13%), found potential drug targets, and enrichment in microglia and three neuronal cell types.
More details below ⬇️

Genomic analyses reveal new insights into Alzheimer's disease Alzheimer's disease (AD) is the most common cause of dementia, with global case numbers projected to reach 153 million in 2050. AD is highly heritable, with twin-based heritability estimates of 60-80%...

🚨 The preprint for our GWAS of Alzheimer’s disease (AD) is out

Preprint: medrxiv.org/content/10.1...

🧵

A big thank you to Wouter Peyrot for his great supervision and teaching me a great deal about stats gen, and to my other co-authors @daniposthu.bsky.social, Alkes Price, as well as to all members of the schizophrenia and major depressive disorder working groups of the psychiatric genomics consortium

GitHub - euffelmann/bpc: Bayesian polygenic score Probability Conversion (BPC) approach Bayesian polygenic score Probability Conversion (BPC) approach - GitHub - euffelmann/bpc: Bayesian polygenic score Probability Conversion (BPC) approach

Code and a tutorial for the BPC approach can be found here:
github.com/euffelmann/bpc

A limiting factor for the usefulness of the BPC approach is the magnitude of R2. While most PGSs explain little variance, some are already proposed to have clinical utility; as GWAS sample sizes increase, their utility will also grow. See the paper for more limitations: rdcu.be/eIjvC

We show in simulations and empirical data that this simple way of estimating R2 works surprisingly well, outperforming another published approach.

In a population reference sample (e.g., 1000 Genomes), where the sample disorder prevalence is the same as in the population, the variance of a PGS on the liability scale will be equal to its R2. That is, no phenotype data is required.

Our approach depends on a valid estimate of R2. Because we wanted to avoid requiring a tuning dataset that will rarely be available in clinical settings, we developed a new way of estimating R2 using GWAS sumstats and public reference data only.

We also compared the calibration of BPC to other methods using tuning samples (with geno- and phenotype data) and show that it performs similarly at smaller tuning sample sizes, but worse at larger tuning sample sizes. Because tuning samples are difficult to obtain, BPC may often be preferred.

We note that the calibration of the Pain et al (2022) method can be improved with some simple tweaks, which we explore in the supplement. The BPC approach still achieves better calibration.

It is also well calibrated in empirical analyses, where we analyzed 9 disorders of varying genetic architectures

We show in simulations, across different parameter settings, that the BPC approach is very well calibrated, outperforming a published method.

To evaluate the BPC approach, we use a metric called the Integrated Calibration Index (ICI): the weighted average of the absolute difference between the real disorder probability and the predicted disorder probability, where 0 indicates perfect calibration.