Michael Bogdos

Hit to Lead best practice? I'm now in Trinidad and I'll share a 180° panorama from Paramin where I walk for exercise. This district in Trinidad's Northern Range is ren...

Took me a while to read but this post by Peter Kenny is well worth the time.

fbdd-lit.blogspot.com/2026/01/hit-...

Quantitative Structure–Reactivity Relationships for Synthesis Planning: The Benzhydrylium Case Selective and feasible reactions are among the top targets in synthesis planning. Mayr’s approach to quantifying chemical reactivity has greatly facilitated the planning process, but reactivity parameters for new compounds require time-consuming experiments. In the past decade, data-driven modeling has been gaining momentum in the field, as it shows promise in terms of efficient reactivity prediction. However, state-of-the-art models use quantum chemical data as input, which prevent access to real-time planning in organic synthesis. Here, we present a novel data-driven workflow for predicting reactivity parameters of molecules that takes only structural information as input, enabling de facto real-time reactivity predictions. We use the well-understood chemical space of benzhydrylium ions as an example to demonstrate the functionality of our approach and the performance of the resulting quantitative structure–reactivity relationships (QSRRs). Our results suggest that it is straightforward to build low-cost QSRR models that are accurate, interpretable, and transferable to unexplored systems within a given scope of application. Moreover, our QSRR approach suggests that Hammett σ parameters are only approximately additive.

E values and LUMO energy correlate quite well (random example: pubs.acs.org/doi/10.1021/...) so you could get an estimate through that correlation.

Goose chase meme. Panel 1, goose says: confounder of what? Panel 2, goose chases person, yelling: CONFOUNDER OF WHAT

When researchers bring up confounders without ever having declared the actual analysis goal

Congrats to Emilien, Maurus and Giorgia! Metal- and CO-Free Carbonylation of Alkyl Iodides @jacs.acspublications.org pubs.acs.org/doi/abs/10.1...

An Inverse Electron-Demand Diels–Alder Approach to Selective Activity-Based Sensing of Acetaldehyde in Living Cells Acetaldehyde (AA) is a reactive aldehyde primarily produced in cells as a metabolic intermediate during ethanol oxidation. Excess AA, often resulting from impaired AA detoxification, leads to aberrant...

Excited to share work out in @jacs.acspublications.org led by Yuxuan and @genlichem.bsky.social on activity-based sensing of acetaldehyde using an inverse electron-demand Diels-Alder reaction, enabling selective detection of two-carbon metabolism! pubs.acs.org/doi/10.1021/...

Congrats to @mixalhsmpogdos.bsky.social, @sevenroediger.bsky.social, @fruepp.bsky.social, Nathalie, Patrick, Fabio and Jan on this epic study of C-N red. elim. in @jacs.acspublications.org - using causal inference, organometallics, electrochemistry, kinetics and DFT. pubs.acs.org/doi/full/10....

Also thanks to @robpollice.mstdn.science.ap.brid.gy for sharing our paper!

bsky.app/profile/robp...

If we managed to map the causal structure aka "the data generating process", we are on track to refine the models.

I believe what we presented is approaching that; imprecisions remain e.g. treating CN as exposure and not modeling S -> CN. Hopefully these are eventually resolved.

8/

Sage Journals: Discover world-class research Subscription and open access journals from Sage, the world's leading independent academic publisher.

As one of my favourite papers puts it:

> As a consequence, linear regression results are notoriously unstable – even minor changes in model specification can lead to coefficient estimates that bounce around like a box full of gerbils on methamphetamines.

journals.sagepub.com/doi/10.1177/...

6/

The biggest takeaway from the paper is not the predictive model we use to infer generalisability, but rather the model structure. The coefficients may change if e.g. we find better descriptors. Even if the model is "retrained" in the lingo of data-driven models, the causal structure remains.

5/

Causal inference lets us visualise, quantify and compare effects. Prior to now, there was no answer to the question "what affects RE most, ligand sterics, electronics, or coordination number?".

Our analysis indicates that electronics is dominant, particularly when changed by lowering CN.

4/

Instead, we propose that CN affects electronics; for CN (2n - 1) fewer donating ligands surround the metal compared to CN (2n), making it more e-poor accelerating RE.

The relevant relationships for variation only in the ancillary ligand(s) are captured in a DAG:

3/

Contrary to textbooks, our data don't support an effect of coordination number on rate, once we condition for metal electronics.

i.e. if we ensure that the metal complexes we compare are equally electron-poor, they eliminate at the same rate, even if one is square-planar and one is T-shaped.

2/

Great to see this published!

In short, we introduced causal inference logic to test which of metal electronics, ancillary ligand sterics and metal coordination number (CN) directly affect the rate of reductive elimination at Pd(II) (and presumably other metals...)

1/

Pleased to have been awarded the ETH Medal this summer for my PhD thesis!

Many thanks to all those who supported me through my journey, especially my PhD supervisor Prof. Bill Morandi.

Congrats to @fruepp.bsky.social, Vasily Grebennikov, Mykola Avramenko, Marc-Olivier Ebert "Kinetic, Spectroscopic, and Computational Investigation of Oxidative Aminative Alkene Cleavage Reveals an N-Iodonium-Iminoiodinane Pathway" now @chemrxiv.org - chemrxiv.org/engage/chemr...

Nature Catalysis, my favourite physical chemistry journal

We have two open PhD positions to start our new group at University of Geneva!
If you are interested in organometallic chemistry, heterometallic complexes and mechanistic investigations, please contact me. Check more details here:
jobs.unige.ch/www/wd_porta...

Thanks for the suggested reading

Could you please walk me through how this works? I've been trying to understand this argument for ages and cannot find sources with hard numbers. It implies that, excluding marketing costs (which are huge), companies are selling at a loss in all other places which I simply cannot find evidence for.

Blunt letter to the editor on ACS: "If ACS won’t stand up now, it should stand down." cen.acs.org/business/Rea... #chemsky 🧪⚗️

100 %.

Very frustrating that EU institutions increase funding when they eye US researchers, but don't care about scientists in Europe who leave science for other careers in droves. Why endorse American exceptionalism?

That is to say nothing of the intra EU inequality, a whole other can of worms.

Pharma CEOS Speaking Up, Damn It

Re-upping this post this morning: Big Pharma CEOs are finally starting to speak out about the Trump administration.

Unfortunately, they’re smiling and clapping.

Stand Up And Be Counted

Biopharma companies and CEOs are keeping their heads down at their own peril. They should speak up about what’s happening to the NIH and other science agencies before it’s too late.

Silence gives consent. And no one should consent to this.

When the federal government abandons important areas of funding and leaves a gaping hole in the advancement of science, any prudent funder would fill the hole, not follow suit. I don’t know the rationale here but seriously WTF?

Disclosure: HHMI paid my way through graduate school.

Hey American PIs. If you have undergrad students who, for whatever personal reasons might find it...appropriate to not do grad school in the good ol' US of A for the next 4+ years, I (and my colleagues at UWindsor Chem) are open to applicants. They pay Canadian domestic rates at UWindsor.

And you WON'T BELIEVE the incentives the Canadian Government will give you to just do R&D in Canada. It's literally insane. Anyways, happy to talk and connect people if anyone is interested. Things might get rocky for a bit, and staff might be happier living somewhere that thinks they are humans.

How does one include all sources of error for a quantity of interest obtained by fitting a model e.g. rate constants from kinetics?

Simulations give the answer:

atigdnc.netlify.app/post/error_handling/

tldr; fit each dataset, propagate and combine errors (or random effects model). #chemsky

Wrapping up 2024 - quite the year! I defended my PhD, after more than 4 years at ETH.

Here's hoping 2025 has something just as exciting in store!

Experimental results here:

foundry.adaptyvbio.com/competition