Na Cai

Na Cai

@caina89.bsky.social

Assistant Prof at D-BSSE, ETH Zurich, studying genetics of complex traits, with a focus on psychiatric disorders www.nacailab.com

929 followers191 following84 posts

Na Cai
@caina89.bsky.social

Both Yazheng and I have learnt a lot from working on this project. We thank the participants of UKB and other cohorts for enabling our work, and many friends for giving us valuable feedback. Yazheng will be presenting this work at #recomb26 @recombconf.bsky.social 18/n

Na Cai
@caina89.bsky.social

Overall, our EDGAR framework enables the prediction of disease-specific liabilities from EHR events, disease-specific measures and deep phenotype labels, and the separation of genetic effects on disease-specific liability from heritable biases that influence EHR events. 16/n

Na Cai
@caina89.bsky.social

We hope this work provides a proof of concept and blueprint for improving the specificity and interpretability of EHR-based genetic studies, as well as their downstream utility in drug target identification and risk stratification. 17/n

Na Cai
@caina89.bsky.social

Finally, we ask if the Common Bias identified in UKB is likely generalizable between EHRs: consistent with our hypothesis, we find that it has high rGs with external EHR GWAS, but not external deep phenotype GWAS. This makes us attempt removing it from external EHR GWAS. 14/n

Na Cai
@caina89.bsky.social

When we did that, we find that high rGs between external EHR GWAS with confounding traits disappeared. This implies that when we have well-predicted disease liabilities, enabled by EDGAR, we can identify biases in one EHR, and remove it from existing GWAS on another. 15/n

Na Cai
@caina89.bsky.social

This drives us to first isolate the disease-specific heritable confounder affecting EHR codes of each disease, using GWAS-by-subtraction, and find that genetic effects contributing to these biases in different diseases have high rGs (i.e. they are shared) with each other. 12/n

Na Cai
@caina89.bsky.social

In contrast, raw UKB EHR codes have lower rGs with external deep phenotype GWAS, but higher rGs and hits replication in external EHR GWAS - this shows replicability between EHRs goes beyond etiological nature of diseases - it is likely an EHR-systemic factor drives this. 10/n

Na Cai
@caina89.bsky.social

We then perform and compare GWAS on raw EHR codes, deep disease labels and EDGAR predicted liabilities: for most diseases, EDGAR predictions give more GWAS hits, higher rGs with deep phenotype labels, and higher PRS predictability and specificity. 8/n

Na Cai
@caina89.bsky.social

We validate our findings in external GWAS: compared to EHR phenotypes, EDGAR liabilities show higher GWAS hit replication and rGs with external deep phenotype GWAS - this suggests EDGAR liabilities capture replicable, disease-specific genetic effects. 9/n

Na Cai
@caina89.bsky.social

We augment EDGAR with an active learning model that prioritizes individuals for obtaining deep disease labels, which reduces the N labels needed by >50% - substantially reducing costs for obtaining disease labels through patient recall in realistic EHR settings. 6/n

Na Cai
@caina89.bsky.social

We first show that raw EHR codes have low correlations with deep disease labels for nine diseases in UKB, and EDGAR liability predictions do significantly better, especially when incorporating disease-relevant measures as inputs. 7/n

Na Cai
@caina89.bsky.social

In our new paper, we propose EDGAR (EHR Disease liability prediction for Genetic Architecture Recovery), a prediction framework that combines the scale of EHR with the disease-relevance of deep phenotypes to break this circularity of bias. 4/n

Na Cai
@caina89.bsky.social

EDGAR predicts lifetime disease liabilities through aligning counts of EHR events with disease-specific measures realistically available in EHRs (e.g. blood biochemistry or spirometry) and independently ascertained, clinically validated disease labels (“deep phenotypes”). 5/n

Na Cai
@caina89.bsky.social

First some background: using EHR codes in GWAS drives increase in statistical power in GWAS meta-analyses. Its potential, however, is undermined by systemic factors like coding practices and healthcare-access disparities between demographics: www.nature.com/articles/s41... 2/n

nature.comHarnessing EHR data for health research - Nature MedicineElectronic health records hold immense potential for providing clinically useful insights for populations and individuals; this Review summarizes the opportunities and challenges, with an emphasis on ...
Na Cai
@caina89.bsky.social

These influences, many of which hertiable, are difficult to disentangle from disease-specific genetics. They may also be replicable across EHRs, making confounded genetic findings look robust. Finally, these biases are likely to propagate in next-event prediction models. 3/n

Na Cai reposted

Jonathan Pritchard
@jkpritch.bsky.social

New preprint alert: we use sign errors as a test of how well TWAS works. Very worryingly we find that TWAS gets the sign wrong around 1/3 of the time (compared to 50% for pure guessing). You can read more about our analysis here, and what we think is going on 👇

Nikhil Milind@nikhilmilind.dev

How well does TWAS estimate a gene’s direction of effect on a trait? We think of this as an important stress-test for the accuracy of TWAS. In a new pre-print, we find that TWAS gets the sign wrong around 20-30% of the time! doi.org/10.64898/202... 1/n

doi.orgHigh false sign rates in transcriptome-wide association studiesTranscriptome-wide association studies (TWAS) are widely used to identify genes involved in complex traits and to infer the direction of gene effects on traits. However, despite their popularity, it r...

Na Cai reposted

Rayan Chikhi
@rayanchikhi.bsky.social

🌎👩‍🔬 For 15+ years biology has accumulated petabytes (million gigabytes) of🧬DNA sequencing data🧬 from the far reaches of our planet.🦠🍄🌵 Logan now democratizes efficient access to the world’s most comprehensive genetics dataset. Free and open. doi.org/10.1101/2024...

Na Cai reposted

Peter Sudmant
@psudmant.bsky.social

The Sudmant lab at UC Berkeley is seeking a postdoc to work on a fully funded NIH project to understand differences in DNA repair and somatic mutation across the primate tree of life. Please spread widely to those who may be interested aprecruit.berkeley.edu/JPF05052

aprecruit.berkeley.eduPostdoctoral Scholar – Genomics, Aging, Somatic Mutation, Structural Variation, Evolution , Cancer – Integrative BiologyUniversity of California, Berkeley is hiring. Apply now!
View full thread
View full thread
Na Cai
@caina89.bsky.social

Though our investigations and findings are centered on psychiatric disorders, the implications are generalizable to all complex traits and diseases, especially those with heterogeneous architectures and unclear diagnostic boundaries. 11/n

View full thread
Na Cai
@caina89.bsky.social

Finally, we find that common genetic effects across all five psych disorders, expected to capture common etiological axes among them, forms a cross-order set most plausibly explained by common confounders external to each disorder’s etiology. 9/n

Na Cai
@caina89.bsky.social

Overall, our work provides a novel metric, the CE test, for informing diagnostic boundaries. Our results show that genetic effects on psych disorders act in sets, and calls for a re-evaluation of current approaches and assumptions. 10/n

View full thread
Na Cai
@caina89.bsky.social

These findings imply that rGs are insufficient to inform nosology, and that the CE test provides a way to determine whether putative disorders or subtypes should be merged or remain separate diagnostically. 7/n

Na Cai
@caina89.bsky.social

We further show that disorder-specific sets lead to comorbidity between two disorders, refuting the assumption that comorbidity is the result of additive effects of genetic effects on both, exemplified by the expectation that high rGs would lead to high comorbidity. 8/n

View full thread
Na Cai
@caina89.bsky.social

In our paper, we test for CE in five psychiatric disorders in the @ukbiobank.bsky.social and the iPSYCH dataset using both polygenic risk scores (PRS) and family-based genetic risk scores (FGRS). We find negative CE for 3 out of 5 disorders, demonstrating there are unrecognized subtypes. 5/n

Na Cai
@caina89.bsky.social

We also show that sets of genetic effects are disorder-specific, despite high genetic correlations (rG) between them. This opposes a well-established conjecture, based on high rG, that clinical boundaries do not reflect their pathogenic processes. 6/n

View full thread
Na Cai
@caina89.bsky.social

These sets, should they exist, would each drive an unrecognized etiological subtype. Our paper tests for the existence of these sets in five psychiatric disorders. 3/n

Na Cai
@caina89.bsky.social

The existence of these sets induces a structured form of statistical interactions called coordinated epistasis (CE), detailed in previous publications by Andy Dahl, Noah Zaitlen www.pnas.org/doi/10.1073/...; negative CE is seen when there are different sets of genetic effects. 4/n

pnas.orgA model and test for coordinated polygenic epistasis in complex traits | PNASInteractions between genetic variants—epistasis—is pervasive in model systems and can profoundly impact evolutionary adaption, population disease d...
Na Cai
@caina89.bsky.social

Our paper addresses a central question in quantgen: whether all genetic variants associated with a complex disease contribute additively to its liability as commonly assumed, or there exists sets of them that co-occur more in certain cases than expected under additivity. 2/n