Lightnews — Scholar-powered news

Emil Uffelmann @euffelmann.bsky.social · 11d

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

Emil Uffelmann @euffelmann.bsky.social · 11d

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

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

github.com

1 1

Emil Uffelmann @euffelmann.bsky.social · 11d

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

1 2

Emil Uffelmann @euffelmann.bsky.social · 11d

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

1

Emil Uffelmann @euffelmann.bsky.social · 11d

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.

1 1

Emil Uffelmann @euffelmann.bsky.social · 11d

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.

1 1

Emil Uffelmann @euffelmann.bsky.social · 11d

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.

1

Emil Uffelmann @euffelmann.bsky.social · 11d

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.

1

Emil Uffelmann @euffelmann.bsky.social · 11d

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

1

Emil Uffelmann @euffelmann.bsky.social · 11d

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

1

Emil Uffelmann @euffelmann.bsky.social · 11d

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.

1

Emil Uffelmann @euffelmann.bsky.social · 11d

This is achieved by transforming a Bayesian PGS (computed using an existing method, e.g., PRS-CS or SBayesR) to its underlying liability scale, estimating the variances of the PGS in cases and controls based on theory, and applying Bayes’ Theorem to compute the probability.

1 1

Emil Uffelmann @euffelmann.bsky.social · 11d

With minimal input (i.e., mostly publicly available data + a prior), the BPC approach transforms an individual's PGS into a probability. No tuning data set with both geno- and phenotype data is required.

1

Emil Uffelmann @euffelmann.bsky.social · 11d

To make PGSs directly interpretable for single individuals, we developed the Bayesian polygenic score Probability Conversion (BPC) approach.

1

Emil Uffelmann @euffelmann.bsky.social · 11d

Accordingly, PGSs are mostly only evaluated using “group-level” metrics, such as the coefficient of determination (R2). A single PGS by itself is more or less meaningless.

1

Emil Uffelmann @euffelmann.bsky.social · 11d

A PGS summarizes an individual’s genetic risk for a disorder in a single value. This PGS, however, is commonly only interpretable if compared to a distribution of PGSs (e.g., a PGS falls in the top 5% of a given distribution).

1

Emil Uffelmann @euffelmann.bsky.social · 11d

Our new paper is out, in which we developed an approach to transform Polygenic Scores (PGSs) into disorder probabilities (i.e., the absolute lifetime disorder risk).

Below a thread 👇

open access link: rdcu.be/eIjvC

Estimating disorder probability based on polygenic prediction using the BPC approach

Nature Communications - Here the authors present a method to transform polygenic scores into disorder probabilities using only GWAS summary statistics, genotype data and a prior - no tuning sample...

rdcu.be

1 6 19

Reposted by Emil Uffelmann

Michel Nivard @michelnivard.bsky.social · 19d

The US as viewed by latenight comedians in Europe (it’s 20 secs of Dutch, the rest is English. We are so worried about you all we’re specifically trying to reach you through our latenight I guess…)

Arjen Lubach @arjenlubach.bsky.social · 19d

www.youtube.com/watch?v=7vRX...

Jimmy Kimmel’s show has been cancelled | The New Disney | LUBACH

YouTube video by LUBACH

www.youtube.com

1 12 12

Reposted by Emil Uffelmann

Our World in Data @ourworldindata.org · 27d

Technology can change the world in ways that are unimaginable until they happen.

Switching on an electric light would have been unimaginable for our medieval ancestors. In their childhood, our grandparents would have struggled to imagine a world connected by smartphones and the Internet.

This image presents a long-term timeline of technology from the distant past to the present and into the future. The timeline is divided into several sections.

At the top, there's a linear progression highlighting significant technological advancements, starting from around 1800. Key milestones include the invention of the steam locomotive, the first vaccine, and the discovery of DNA. The timeline shows notable events like the Wright brothers' flight in 1903, the beginning of the Internet in 1991, and the 21st century’s focus on artificial intelligence and space exploration.

The timeline also includes a spiral that represents the vast expanse of human history, where each turn symbolizes 200,000 years. This section mentions major milestones like the use of tools, the control of fire, and the emergence of Homo sapiens.

The image is labeled "A long-term timeline of technology" and is attributed to Our World in Data, created by Max Roser, licensed under CC-BY.

2 15 55

Reposted by Emil Uffelmann

Jonathan Sebat @sebatlab.bsky.social · Jul 16

As we have learned, genes have dose-dependent effects on psychiatric traits. DOSAGE, it turns out, is a key element that helps unravel mechanisms of gene → pathway → cell type → brain region → diagnosis. Here we developed a framework to characterize cellular processes that mediate genetic effects.

medRxivpreprint @medrxivpreprint.bsky.social · Jul 14

Psychiatric disorders converge on common pathways but diverge in cellular context, spatial distribution, and directionality of genetic effects https://www.medrxiv.org/content/10.1101/2025.07.11.25331381v1

2 8 30

Emil Uffelmann @euffelmann.bsky.social · Aug 23

Thanks! For intelligence, there was a relatively small number of loci for which we could compute rgs (due to low h2 in both or one sex). So I wouldn’t interpret the distribution too much. For neuroticism, perhaps something similar (need to check), but could suggest more GxS than other traits

1

Reposted by Emil Uffelmann

Sasha Gusev @sashagusevposts.bsky.social · Aug 10

I wrote about how genetic risk works in the context of embryo selection and how people often think about it all wrong. A short 🧵:

What we talk about when we talk about risk

How embryo selection exploits our flawed intuitions about risk

open.substack.com

3 30 75

Emil Uffelmann @euffelmann.bsky.social · Aug 6

Many more results can be found in the paper: rdcu.be/ezuZd.

7/7

Local genetic sex differences in quantitative traits

Nature Communications - Analysing 157 traits, this study finds widespread local genetic sex differences masked at the genome-wide level. Using LAVA, it tests for sex-specific heritability, genetic...

rdcu.be

Emil Uffelmann @euffelmann.bsky.social · Aug 6

E.g., when we applied this test to gene boundaries, we found APOE showed substantial differences for LDL, with larger effects and h² in females (i.e., 6% vs. 3%)

6/7

1 1

Emil Uffelmann @euffelmann.bsky.social · Aug 6

We not only tested for equality of direction of genetic effects and their relative magnitudes (i.e., correlations) but also for equality of their absolute magnitude. Nearly all traits had at least one locus that differed between males and females using this test.

5/7

1