iiinova — where your digital self keeps evolving

Mykhaylo M. Malakhov @mykmal.xyz · 04/10/2026

I'm more worried about AI eroding job desirability than eliminating jobs. As software engineering and statistics roles increasingly turn into prompting exercises, we will see fewer young people choosing them -- even though expertise in these fields is essential for understanding and managing AI.

Dariia Mykhailyshyna @handle.invalid · 19/09/2026

Reposted by Mykhaylo M. Malakhov

Looking for more speakers for Workshops for Ukraine (sites.google.com/view/dariia-...), so if you want to contribute to a series by teaching a workshop on #Rstats, #Python, data analysis, while helping raise money for Ukrainian charities, drop me a message or email dmykhailyshyna@kse.org.ua #Econsky

01616

Reposted by Mykhaylo M. Malakhov

Joshua G. Schraiber @jgschraiber.bsky.social · 08/09/2026

This is fucking insane. It's unclear what the future of mathematics looks like at this point. It's hard to express how big of an advance this is. Yes, there are serious issues about credit here, but all parties agree: this proof was enabled and perhaps principally shaped by AI.

2338

Mykhaylo M. Malakhov @mykmal.xyz · 04/09/2026

I'm no prophet, but I did write a 15-page essay in high school arguing that tech companies should embrace open source because the biggest revenue streams in the future will be services and compute, not code.

Josh Weinstock @joshweinstock.bsky.social · 25/08/2026

Reposted by Mykhaylo M. Malakhov

Excited to share our latest foray into studying the ways in which germline variation gives rise to somatic variation in aging blood: www.medrxiv.org/content/10.6... In work led by @prarthana-s-r.bsky.social , we map how common and rare germline variation links to (somatic) passenger mutations 🧬🧵👇

medrxiv.org

Multi-ancestry analysis of 791K whole genomes reveals the genetic, geographic, and phenotypic correlates of somatic passenger mutations in blood

Clonal hematopoiesis (CH), an aging-related expansion of hematopoietic stem cell (HSC) clones, is associated with hematologic malignancy, cardiovascular disease, and mortality. Most clonal expansions,...

176

Mykhaylo M. Malakhov @mykmal.xyz · 25/08/2026

The Ukrainian Biological Data Science Summer School is such a great idea, and I'm glad to see that it is still going strong after four years! I would love to help out in a future iteration.

Mykhaylo M. Malakhov @mykmal.xyz · 19/08/2026

February 2009 was over 17 years ago, folks.

Mykhaylo M. Malakhov @mykmal.xyz · 18/08/2026

Depending on your perspective, that's either a tragedy or an opportunity to take a vacation for the rest of the week 😅

Mykhaylo M. Malakhov @mykmal.xyz · 18/08/2026

Already? It's not even 9am yet. I guess I need to step up my game.

Mykhaylo M. Malakhov @mykmal.xyz · 18/08/2026

This is all I have to show for my data harmonization efforts, sorry.

130

Mykhaylo M. Malakhov @mykmal.xyz · 13/08/2026

People who completed their education right when capable LLMs were becoming available will be the biggest beneficiaries of AI-driven productivity gains. Everyone before them had to build their careers without the help of AI, while those after them are missing out on the rigor of pre-AI learning.

030

Mykhaylo M. Malakhov @mykmal.xyz · 15/07/2026

Very interesting! LLMs seem to be quite good at finding pathological counterexamples to conjectures that otherwise hold in the more typical settings humans tend to consider.

Mykhaylo M. Malakhov @mykmal.xyz · 15/07/2026

This is very cool work! Most research in statistical genetics does not take evolutionary theory into account, but Hanbin's paper shows how doing so can explain the success of empirically-derived models and make them even more effective.

030

Mykhaylo M. Malakhov @mykmal.xyz · 15/06/2026

I am grateful to the Center for Computational, Evolutionary and Human Genomics for its support and am looking forward to engaging in cross-disciplinary conversations with other faculty and trainees as a CEHG Fellow!

Mykhaylo M. Malakhov @mykmal.xyz · 15/06/2026

This fellowship will partially fund the second year of my postdoctoral training at Stanford, where I will continue working with @johnwitte.bsky.social and @lindakachuri.bsky.social to develop personalized biomarkers for prostate cancer and other complex diseases.

120

Mykhaylo M. Malakhov @mykmal.xyz · 02/06/2026

"mtDNA copy number and expression are largely decoupled, and above a certain rate-limiting threshold, mtDNA accessibility — rather than absolute copy number — is the more relevant quantity for explaining inter-individual differences in gene expression."

011

Mykhaylo M. Malakhov @mykmal.xyz · 29/05/2026

IGES is my favorite scientific conference. It has a small, intimate feel and remains focused on genetic epidemiology and statistical genetics, which makes networking easy and puts your research in front of an enthusiastic audience. I'm going again and hope to see many of you there!

022

Mykhaylo M. Malakhov @mykmal.xyz · 06/05/2026

After closely following generative AI research for the past several years but intentionally not using AI in my own work, I finally decided to bite the bullet. So far, it is both more useful and less useful than I expected.

120

Mykhaylo M. Malakhov @mykmal.xyz · 28/04/2026

To me, the most compelling argument against AI in this setting is the observation that the process of scientific research and writing is inherently valuable, regardless of its output. @minaskar.bsky.social expressed it eloquently in the essay I quote from here: bsky.app/profile/mykm...

Mykhaylo M. Malakhov @mykmal.xyz · 08/04/2026

This might not be a smart policy decision. Yesterday I went to a talk on LLM resilience to adversarial input, and it turns out that lower-reasoning versions of GPT-5 are more likely to suggest wrong diagnoses after they are given distracting input. Here's the preprint: arxiv.org/pdf/2603.14158

arxiv.org

Mykhaylo M. Malakhov @mykmal.xyz · 04/04/2026

"What matters is the process: the development and application of methods, the training of minds, the creation of people who know how to think about hard problems. If you hand that process to a machine, you haven't accelerated science. You've removed the only part of it that anyone actually needed."

Mykhaylo M. Malakhov @mykmal.xyz · 01/04/2026

Where did you find longitudinal data on time to acceptance? At most I have only seen journals provide a single estimate, if they publish this information at all.

100

Mykhaylo M. Malakhov @mykmal.xyz · 25/03/2026

It's so exciting to see the company where I interned as a PhD student obtain their first FDA approval! This is huge news for the Hunter Syndrome community and a major milestone in the development of biotherapeutics that cross the blood-brain barrier. investors.denalitherapeutics.com/news-release...

investors.denalitherapeutics.com

Denali Therapeutics Announces U.S. FDA Approval of AVLAYAH™ (tividenofusp alfa-eknm) for Treatment of Hunter Syndrome (MPS II) | Denali Therapeutics

The Investor Relations website contains information about Denali Therapeutics's business for stockholders, potential investors, and financial analysts.

011

Mykhaylo M. Malakhov @mykmal.xyz · 23/03/2026

Congrats, it's exciting to see the published version of this work!

011

Mykhaylo M. Malakhov @mykmal.xyz · 19/03/2026

Coming from a background in math, I used to hate the results-first format. Now I think that both are appropriate in different contexts. If a paper proposes a new method, I want to see it described first. But if the focus is on answering a biological question, putting the results first is better.

Mykhaylo M. Malakhov @mykmal.xyz · 18/03/2026

I'm proud to see that my PI @lindakachuri.bsky.social has been elected to the @geneticepi.bsky.social Board of Directors! It's time to make IGES cool again 😎

Mykhaylo M. Malakhov @mykmal.xyz · 11/03/2026

It would have been even better if more people dropped out before the filing deadline. Assuming they all qualify, we will see eight Democrats on the ballot even if some drop out before the election.

121

Mykhaylo M. Malakhov @mykmal.xyz · 25/01/2026

Thanks, those are good suggestions. I appreciate your words of encouragement.

Mykhaylo M. Malakhov @mykmal.xyz · 23/01/2026

Those issues are the most personal for me, but they are just the tip of the iceberg among the long list of things that are seemingly falling apart in the world. How is it possible to stay sane and continue focusing on research when so much else is at stake?

050

Mykhaylo M. Malakhov @mykmal.xyz · 23/01/2026

I wanted to keep this account focused on science, but it is hard to stay silent on political issues right now. My home country is still being bombed by Russia, the city I lived in for the past several years is under occupation by ICE, and the funding that supports my work is constantly threatened.

260

🧵11/11 Finally, I want to extend a big thank you to Wei Pan for being my dissertation advisor and guiding me throughout this project. I am also grateful to @zkutalik.bsky.social and Guimin Gao for serving as peer reviewers and providing feedback that significantly strengthened this paper.

000

GitHub - mykmal/cowas: Association testing for genetically regulated co-expression.

🧵10/11 Interested in trying out COWAS yourself? Our code is available at github.com/mykmal/cowas and weights for all models trained in this paper are provided at www.synapse.org/cowas.

github.com

Association testing for genetically regulated co-expression. - mykmal/cowas

100

🧵9/11 By leveraging information from multiple proteins, COWAS also boosts power relative to standard PWAS. For example, alpha-synuclein (SNCA) did not attain significance for Parkinson's when tested on its own, but became significant once considered jointly with any one of several other proteins.

100

🧵8/11 COWAS can also help to disentangle direct and interaction effects. For example, we found that APOE and its genetically regulated interaction with LDLR have opposite effects on Alzheimer's disease, which is consistent with known APOE biology. Standard PWAS was only able to identify APOE itself.

100

🧵7/11 Overall, COWAS uncovered dozens of interacting protein pairs with significant effects on disease risk. We identified many signals missed by standard PWAS, showing that co-expression may affect complex traits even if neither protein is detected to influence the trait when considered on its own.

100

🧵6/11 We applied COWAS to @ukbiobank.bsky.social proteomic data and the largest GWASs for LDL cholesterol, Alzheimer's disease, and Parkinson's disease. Although co-expression models had lower performance than single-protein models, we still obtained quality models for over 500 pairs for each trait.

100

🧵5/11 Next, COWAS tests for association between genetically regulated co-expression and the trait of interest while also accounting for direct effects from each exposure. This joint model can be estimated using GWAS summary statistics, making our method just as easy to use as standard TWAS and PWAS.

🧵4/11 We developed the COWAS framework to answer this question. First, COWAS trains models to predict expression and co-expression from common genetic variants. In our main results we defined co-expression as the product of expression residuals, removing the linear genetic components of expression.

100

🧵3/11 This assumption can be a good approximation, but it ignores the effects of gene–gene or protein–protein interactions. Recent research has shown that interaction and co-expression effects are regulated by genetic variation. But how does genetically regulated co-expression affect complex traits?

🧵2/11 Most methods for identifying disease-relevant genes or proteins only consider one exposure at a time. TWAS trains gene-level models to predict expression from genetic variation and then tests each gene's imputed expression for association with an outcome trait. PWAS does the same for proteins.

Co-expression-wide association studies link genetically regulated interactions with complex traits - Nature Communications

My second dissertation paper is now out in @natcomms.nature.com! Here we introduce the co-expression-wide association study (COWAS) method for identifying pairs of genes or proteins whose genetically regulated co-expression is associated with complex traits. 🧵⬇️ www.nature.com/articles/s41...

nature.com

Proteins can affect traits through their interactions with one another. Here, the authors propose the COWAS method to identify pairs of interacting genes or proteins whose genetically regulated co-exp...

1145

Mykhaylo M. Malakhov @mykmal.xyz · 11/12/2025

Now we know what those sky-high APCs are being used for...

Mykhaylo M. Malakhov @mykmal.xyz · 23/11/2025

"The aim of equity is not to brag about how many students of a target background you awarded an A in calculus, it’s to ensure that every student actually learns." This is such an important and timely article. I have noticed a lot of the same issues myself. www.theargumentmag.com/p/when-grade...

theargumentmag.com

When grades stop meaning anything

The UC San Diego math scandal is a warning

Mykhaylo M. Malakhov @mykmal.xyz · 08/11/2025

Me too, especially if the paper has any mathematical notation in it.

Mykhaylo M. Malakhov @mykmal.xyz · 07/11/2025

This is an exceedingly late post, but I wanted to share a few pictures from #IGES2025 earlier this fall. I had the honor of giving a platform talk on age-related genetic dysregulation, and I really enjoyed connecting with other researchers and exploring the beautiful city of Cologne!

030

Mykhaylo M. Malakhov @mykmal.xyz · 07/11/2025

More votes have been counted since then, and the updated map looks even better. There's not a single red cone across the whole state!

Mykhaylo M. Malakhov @mykmal.xyz · 24/09/2025

Reading papers detracts from time that could be spent actually analyzing data, and yet keeping up with new research is necessary to analyze data effectively. I'm still struggling to find a good balance, especially when I feel like I barely have enough time to read my email...