Brandon Logeman

Excited to share VIPerturb-seq! New tech from my lab which aims to improve the cost, data quality, and efficiency of single-cell CRISPR screens so that they are accessible to any lab - even at genome-wide scale Preprint and 🧵 (1/): biorxiv.org/content/10.648…

The latest issue of COB describes how AI infrastructure can create value for biotechnology beyond finding new drug candidates. Highly recommend reading if you work at the intersection of tech and life sciences

@LaramieDuncan @PGCgenetics @SashaGusevPosts @dgmacarthur @tuuliel @hailianghuang Thanks @LaramieDuncan , I'll look up those works. BTW I'm very excited to dig into your recent paper on cell-type associations in complex brain phenotypes...its next on my list!

@brandon_logeman @PGCgenetics @SashaGusevPosts @dgmacarthur @tuuliel @hailianghuang has some excellent publications examining this. It seems to be the exception, more than the rule. Interestingly, this also implies that most risk variants are ancestrally old.

Naive population genetics question: in multi-ancestry GWAS, are ancestry-specific associations common? For example, the recent BD analysis from @PGCgenetics found EAS specific associations. Rule or exception? @SashaGusevPosts @dgmacarthur @tuuliel nature.com/articles/s4158…

@doctorveera @sanogenetics Gotcha, thanks @doctorveera for the explanation that what is being observed is more statistical, rather than biological, in nature.

Thank you for your feedback on the podcasts. Regarding your q, in multi-ancestry GWAS, it's common for authors to also perform ancestry-specific GWAS, which was the case here. The EAS specific locus was from EAS specific GWAS meta-analysis. Note, the ancestry-specific signal doesn't mean ancestry-specific biology. It's simply the signal is detectable in one ancestry but not the others because of allele frequency differences. Should there be a different variant in the same gene present at higher frequency in, for e.g., Europeans, then we will see a similar association in Europeans as well.

@SashaGusevPosts @PGCgenetics @dgmacarthur @tuuliel “…apparent cross-population differences in genetic effects may be largely explained by differences in allele frequencies or linkage disequilibrium with “tagging” variants rather than differences in the underlying effect sizes.” Thanks for the great write up @SashaGusevPosts

@brandon_logeman @PGCgenetics @dgmacarthur @tuuliel Here's a summary of how this looks in the MVP (theinfinitesimal.substack.com/p/minimal-evid…). In general effects are highly shared and even statistical specificity is not always evidence of biological specificity. But there are exceptions.

Excellent write up by @doctorveera found here:gwasstories.com/p/de-novo-enha…

A new paper has identified an answer to this question, in which a structural variant causes the de novo creation of a cardiomyocyte-specific enhancer, leading to ectopic expression of a potassium channel and heart disease.

A big congratulations to our #GWAS24 Poster Winners, Luis Eichelmann (@LuisEichelmann) from @UniLuebeck & Brandon Logeman (@brandon_logeman) from @Harvard 🎉👏

@TonyZador Others have discussed grant funding as a weighted lottery (i.e. better study section score gives you more ping pong balls in the ticket tumbler)....could something similar happen for publications?

@TonyZador Their solution was to inject noise into the ordering of draft picks; a lottery that is weighted by the teams number of wins. It overall allocates resources (draft picks) where they belong but due to individual randomness keeps teams from gaming the system and losing on purpose

Many problems in science a field boil down to Goodhart's law: "When a measure becomes a target, it ceases to be a good measure" Eg a high-profile paper didn't used to be the goal But given finite resources, we need a measure to allocate them, which invites gaming the system

@GeneticsMike7 @BritMartinez Dependent on one's definition of beauty, pretty sure Alaska can give Hawaii a run for its money

@BritMartinez Gotta be Hawaii right?

What’s the most beautiful state in America?

@Jeff_Mold Agreed. I sometimes catch myself thinking the same thing.

@OstuniLab @MolecularCell @pioneerfactors Thanks for highlighting this discussion which has a core topic (advantages / disadvantages of in vitro / in vivo studies) that is relevant beyond pioneer factors and permeates much of biology. Well written on both sides.

Must-read discussion by giants - Ken Zaret and Steve Henikoff - on the nature of "pioneer factors", in vitro vs in vivo, Da Vinci vs Feynman. Learning how chromatin becomes accessible is key for immunology (and biology in general) cell.com/molecular-cell… cell.com/molecular-cell…

@Vale_Tornini Congrats Valerie! 🎉

@GeneticsMike7 @doctorveera @Nature @mnelsonxy @NatureGenet Conceptually the important genes = low effect sizes seems measurable. Practically I'm not sure how it would be done; perhaps some kind of meta study across multiple traits comparing effect size with PhyloP scores or dN/dS...something like that. Anyone tried this?

@brandon_logeman @doctorveera @Nature @mnelsonxy @NatureGenet Eg we know SNCA, APP, MAPT and APOE are central players for PD, AD, etc, but common risk variants at these loci have tiny effects, like other loci. So effect size doesn’t tell you anything about gene importance IMO

I was reading the recent @Nature paper on the impact of human genetics on drug success (nature.com/articles/s4158…). I found one of the findings particularly interesting. ​ This is a follow-up work of @mnelsonxy's famous 2015 paper published in @NatureGenet (x.com/doctorveera/st…). In their 2015 work, Nelson et al. reported that drug programs with human genetic support doubled that clinical success. In the new analysis, @cureffi et al. report that that estimate now increased from 2 to 2.6-fold, mainly because many new genetic discoveries have been published since 2015. ​ Coming to the finding that interested me most. In the paper, the authors address one of the popular criticisms of GWAS, particularly the debate on common vs rare variants. ​ GWASs, based on common variants, typically identify associations with small to tiny effect sizes. Sequencing-based rare variant studies typically identify associations with large effect sizes. The value of common variant associations to drug target discovery or to establishing a causality between a gene and a trait is often questioned. Particularly in the context of explosive growth of GWAS sample sizes, resulting in discoveries of variants with smaller and smaller effect sizes. ​ The authors address this skepticism through stratified analysis and make a couple of interesting observations. ​ Firstly, the authors show that the impact of human genetic evidence on drug success do not statistically differ across different minor allele frequencies or effect sizes or year of publication. Genetic associations of small effect sizes are as much informative as genetic associations of large effect sizes. Note, the above groupings were within the GWAS umbrella. ​ Secondly, when comparing between GWAS source and OMIM (contains Mendelian genetic associations), the authors do find a larger effect size for OMIM compared to GWAS. However, they argue that the relatively larger effect size for OMIM against GWAS sources (such as open targets, GWAS catalog etc.) mainly highlight the current challenges of GWAS to confidently identify causal genes. I've written before on this topic, using FTO as an example (x.com/doctorveera/st…). ​ In line with their hypothesis, the authors find that the impact of GWAS-based human genetic support increase with increase in our confidence on the causality of genes. When stratified based on L2G score (a score developed by @OpenTargets that quantifies the level of evidence available to map a GWAS locus to a gene), the effect sizes increased with increase in L2G score. ​ So, it's not the small or large effect size that is important, but how confident are we in the casual gene at the locus. At least for this reason, the coding variants with large effect sizes will continue to offer better value for drug development than noncoding variants of small effect sizes until we solve the problem of causal gene mapping. ​ I also like that the authors explicitly state what their findings do and don't mean. The findings inform only about how often existing drug programs with genetic support succeed compared to those without. They don't inform anything about the odds of a genetic association to become a successful drug target. That is entirely a different research question. ​ Minikel et al. Nature 2024 nature.com/articles/s4158…

New GWAS study examines when a baby takes their first steps 🧬 ➡️ 🚼 ➡️ 🚶‍♀️ SNP based heritability, a lower bound of heritability calculated using only common variants, shows up at ~25%, much higher than I would have guessed. Work led by @annagui86

@doctorveera @Nature @mnelsonxy @NatureGenet I’m sure folks smarter than me have pondered over this before, if anyone has any thoughts please share; would love to learn more

@doctorveera @Nature @mnelsonxy @NatureGenet But on the other hand I could image a set of coding mutations that very slightly decrease their function but is not enough to be completely eliminated from the population…..in this case perhaps these low effect GWAS genes would be enriched for rare variants in disease cases