Adriaan van der Graaf, PhD
@ajhgnews.bsky.social sat with @adriaan-vd-graaf.bsky.social in the latest "Inside AJHG" to discuss his recently published paper, “Mendelian randomization linking metabolites with enzymes reveals pathway regulation and therapeutic avenues.”➡️ bit.ly/46di69J #ASHG #HumanGenetics
🧬 New preprint alert! After years of collaborative work across 52 datasets we are presenting eQTLGen phase 2: a genome-wide eQTL meta-analysis covering 43,301 blood samples: www.medrxiv.org/content/10.6... (1/8)
Mendelian Randomization is powerful but often debated. We benchmarked it against the "ground truth" of human metabolic pathways to see how often it gets it right. 🧬
Result: MR is robust! It matches known metabolism in 43% of cases and identifies valid drug targets.
www.cell.com/ajhg/fulltex...
My New Year’s resolution had been to write a todo list every day, and print it out.
Add notes to it, and update every morning with yesterdays notes.
Worked great until now.
EMGM 2026 will be in beautiful Davos this year!
emgm2026.sciencesconf.org?lang=en
Registration abstract submission and travel grant applications are open!
Hope to see you there!
Really excited to share our metabolite ratio QTL work Friday!
- 25% more findings from already existing QTL studies
- We proxy enzymatic activity that is otherwise unmeasured
- We implicate ABCG2 as a PFOS 'forever chemical' transporter
Come by the poster and say hi!
Hi #ASHG2025, the Lausanne team made it to Boston ! Excited to present our latest discoveries — come say hi and learn more about our research! @samuelmoix.bsky.social @rjhfmstr.bsky.social
Tip for presentations: avoid saying "I don't think this audience needs an introduction to X". Try instead giving a v brief introduction tailored to the key points you will reference in the rest of the talk.
So often people assume I have knowledge I don't, and this hurts my ability to follow.
Using google gemini with grounding for later reference? Make sure you download your links before they expire... :(
🚨 Our parent-of-origin study is out in Nature! 🧬
Maternal and paternal alleles can have distinct — even opposite — effects on human traits, revealing a hidden layer of genetic architecture that standard GWAS miss.
🔗 www.nature.com/articles/s41...
Highlights below!
Wild ride.
Brain Surfaces of 70 primate species
1
To predict the behaviour of a primate, would you rather base your guess on a closely related species or one with a similar brain shape? We looked at brains & behaviours of 70 species, you’ll be surprised!
🧵Thread on our new preprint with @r3rt0.bsky.social , doi.org/10.1101/2025...
Sure! It’s mostly my own incompetence holding me back! But will do when I encounter a bug or really don’t know what to do!
Congratulations! I’ve been using the code already for some projects! Happy to see it being published!
Super excited to see this out. What started as some math in a grant in 2020, to a student deciding to take this on in 2022, to published in 2025.
These things can take time and patience is key!
I wrote a little bit about a cool recent paper looking at heritability estimates from very large registry data, and how we still really don't understand why outcomes track in families. A short 🧵:
🚨New preprint is out!
How do genetic effects on complex traits change with age? In this work, we compare different approaches to obtain age-varying genetic effects, and show how design and modeling choices can impact the conclusions we draw.
shorturl.at/17snd
A thread 🧵👇
This work wouldn't have been possible without @zkutalik.bsky.social nor Chiara Auwerx, Carolina Borges, Robert Warmerdam, Lude Franke, Urmo Võsa and the eQTLGen consortium!
Thanks for reading all the way, and if you have ANY questions, happy to interact!
So that's MR-link-2 in a nutshell!
There's a lot more little gems in the paper, so do have a read! www.nature.com/articles/s41...
Interested in running MR-link-2? take a look at the GitHub!
github.com/adriaan-vd-g...
All feedback is welcome! Happy to guide even total beginners!
And as a final ground truth we use the markergenes that are used to identify cell types in single cell experiments to the cell type proportions themselves.
Here, the cell type should influence gene expression, and not the other way around! MR-link-2 has very good discriminative ability.
We think MR-link-2 truly estimates pleiotropy!
When we correlate the MR-link-2 estimate of pleiotropy with the deviation (individual Q statistic contribution of each region) from meta-analysis, we find highly significant correlation
When we look at biomarker -- disease causality, we find that MR-link-2 identifies all of the true positives we tested.
All while having a median per locus false positive rate of 0.09 in the true negatives (compared to 0.15 of other methods): MR-link-2 has lower false positive rates
Our ground truth is set up to be quite stringent, but when we look at other significant results outside of the ground truth, MR-link-2 is the only method that finds the causal relationship between pyruvate and citrate, a key citric acid cycle link. As well as other very compelling examples.
And MR-link-2 is very competitive in discriminating between direct reactions and reactions that are super far away (Negative distance).
Here, we find that MR-link-2 is better at identifying the correct causal effect estimate between substrates and products (should be positive) than other MR-methods
Metabolic pathways
We used 4 metabolite QTL studies, and 3 metabolic pathway references to build a network of reactions that an MR method is supposed to detect.
Now, you may be wondering, sounds like something cool, but how ever would you prove that it works?
We spent a lot of time on to build ground truth in human causality. We compiled ground truth causality from metabolic pathways, biomarker-disease pairs and cell types and marker genes.
Diagrams of conventional Mendelian randomization, compared to the MR-link-2 method. The MR-link-2 method uses correlated SNPs in an associated region to estimate a causal effect, and pleiotropy.
The trick is that we use all the SNPs in an associated region, compared to other methods that usually only use one SNP.
We developed MR-link-2, an MR method robust to pleiotropy. The unique bit is that it only requires a single associated region for it to work. Interested?
Have a look at the code and examples on the github!
github.com/adriaan-vd-g...
A network of metabolites and their potential causal relationships. Green edges are Detected by the statistical causal inference method MR-link-2
Our paper, MR-link-2 has just been published! Offering pleiotropy robust Mendelian randomization from a single region! www.nature.com/articles/s41...
