Rowan Research Spotlight: Turki Alturaifi

by Corin Wagen · May 7, 2025

What makes a reaction selective? Can we predict, before doing an experiment, where a catalyst will bind or whether a mutation will improve enzyme activity?

These are the kinds of questions that drive the research of Turki Alturaifi, a fourth-year PhD candidate at the University of Pittsburgh in Professor Peng Liu's group. Originally from Saudi Arabia, Turki was first introduced to computational chemistry as an undergraduate working with Rob Paton at Colorado State University. "There was a seminar by Rob about computational chemistry… that was the first time I heard about it," he said. "What caught my eye was how much money and time you could save—and the chemical waste." From that moment on, he was hooked.

A picture of Turki.

Turki Alturaifi

Reaction Mechanisms, Enzyme Catalysis, and Everything In Between

Turki's current research spans both organic and enzymatic catalysis, unified by a central theme: predicting reactivity and selectivity from first principles. "The general overview," he explains, "is how can we predict reactivities and selectivities in organic reactions?" That often means elucidating reaction mechanisms—especially for challenging systems like alkene functionalizations—and using that insight to design new directing groups, new reactions, or better catalysts.

Some of the most complex systems he's studied are enzymes. "There are a lot of libraries that generate combinatorial mutations," he explains, "and one project I'm working on, in collaboration with Professor Yang Yang, is how can we incorporate quantum chemistry information to give you better scoring." The vision is compelling: combining short molecular dynamics simulations with high-level quantum-chemical calculations to predict how protein dynamics might impact catalytic activity. "We're trying to see if the loop flexibility has any relation to the observed enantioselectivity," he says. "It's very difficult… it's really not straightforward to design these libraries."

Much of the field still relies on sequence-based prediction, but Turki argues that incorporating structure-based insights can make models more useful and more accurate. Starting from AlphaFold models or known structures, his workflow mutates enzymes in silico, runs MD simulations, and attempts to correlate loop behavior with experimental results. It's a computationally demanding approach, but one that reflects the real physical complexity of biocatalysts.

The Heteroaryl Database (HArD): Data at Scale

Turki's also worked on more data-driven projects. "I worked on a heteroaryl database," he said, referring to a dataset of over 30,000 heteroaryl groups for which he and collaborators computed a variety of properties relevant to organic chemistry, like Hammett constants. The goal is simple: "If you're troubleshooting a reaction, how can you quickly obtain these properties?" With this database in hand, researchers can now do just that—query electronic properties of potential substrates and feed that information into predictive models, allowing them to more quickly ascertain substrate trends and build meaningful models of reaction yield and selectivity.

What's Hard About Computational Chemistry?

When asked what makes computational chemistry uniquely difficult, Turki doesn't hesitate. "Accuracy," he says. "[A computational prediction] is not an experimental value. It's a computed value. Is it accurate or not? That alone can be a whole significant work," he explains.

Even if Turki can eventually find a way to obtain accurate predictions, he also wants to gain mechanistic insights. "We always try to know the ‘why,' and that's much harder," he says. Mechanistic inference is a nuanced process, and Turki often has to evaluate five to ten possibilities before he can propose a single pathway.

Considerations like conformational flexibility can make these studies even harder. "Catalysts are getting bigger. It's very difficult," he says. "What looks easy for 50 atoms becomes impossible for 250." In one recent project studying rhodium catalysts (in collaboration with Professors Jamal Musaev and Huw Davies), Turki turned to QM/MM with xTB and molecular dynamics to understand how π-stacking interactions governed selectivity. Initially, the calculations weren't fast—"27 days to run 500 ps," he admits—but they provided valuable insight into the spatial dynamics of these large, flexible systems.

The Role of Rowan

Turki is a fan of how Rowan makes it easy to employ state-of-the-art methods, days or weeks after they're originally published. "I love the innovations," he says, even though he and his group have their own resources and don't primarily use Rowan for their day-to-day research.

He's used Rowan's ML-based pKa model in a paper currently under review, a collaboration with Professor Keary Engle at Scripps. "We wanted to know if this sulfonyl fluoride is acidic enough to be deprotonated," he said. With Rowan's pKa model "you can test it in two seconds, it's perfect." The result helped explain a surprising selectivity in an oxidative-addition reaction, which reviewers questioned during peer review. "The reaction works with sulfonyl fluorides, but not with iodides. We knew from experiment, but this confirmed it."

Turki's pKa calculations in Rowan.

Turki's pKa calculations help to explain surprising selectivity in oxidative addition.

Turki appreciates that Rowan emphasizes transparency and reproducibility: "Rowan gives you optimized workflows you can trust. There are citations, references—it's not like it's pulled from nowhere." And while he's comfortable working in a terminal, he knows that's not true for every chemist. "The difficulty is not opening the terminal—it's the time. As an organic chemist, should you do B3LYP or ωB97X-D? Do you have time to look into this? Rowan gives you the default."

Looking Ahead: A Dream of De Novo Reaction Design

What excites Turki about the future? "The equivalent of de novo protein design for organic molecules," he says. "That's a big dream I have." He hopes for a world where researchers can put in a reaction and receive a predicted substrate scope—or better yet, a predicted catalyst.

But he's skeptical of one-size-fits-all automation. "We don't want to do the decision-making or inference for the scientist," he says. Instead, the goal should be to empower researchers with better tools. "Providing the tools is everything," he says, highlighting the value of software like Rowan.

In the meantime, Turki is keeping busy. He's preparing a revision for Nature Synthesis, working on a follow-up to the heteroaryl database, and building models to predict enzyme activity from structure and quantum descriptors. As the tools improve, and as platforms like Rowan lower the barrier to computational chemistry, Turki's vision of data-rich and mechanism-guided catalyst design gets closer to reality.

Banner background image

What to Read Next

ML Models for Aqueous Solubility, NNP-Predicted Redox Potentials, and More

ML Models for Aqueous Solubility, NNP-Predicted Redox Potentials, and More

the promise & peril of solubility prediction; our approach and models; pH-dependent solubility; testing NNPs for redox potentials; benchmarking opt. methods + NNPs; an FSM case study; intern farewell
Sep 5, 2025 · Eli Mann, Corin Wagen, and Ari Wagen
Machine-Learning Methods for pH-Dependent Aqueous-Solubility Prediction

Machine-Learning Methods for pH-Dependent Aqueous-Solubility Prediction

Prediction of aqueous solubility for unseen organic molecules remains an outstanding and important challenge in computational drug design.
Sep 5, 2025 · Elias L. Mann, Corin C. Wagen
What Isaiah and Sawyer Learned This Summer

What Isaiah and Sawyer Learned This Summer

Reflections from our other two interns on their time at Rowan and what they learned.
Sep 5, 2025 · Isaiah Sippel and Sawyer VanZanten
Benchmarking OMol25-Trained Models on Experimental Reduction-Potential and Electron-Affinity Data

Benchmarking OMol25-Trained Models on Experimental Reduction-Potential and Electron-Affinity Data

We evaluate the ability of neural network potentials (NNPs) trained on OMol25 to predict experimental reduction-potential and electron-affinity values for a variety of main-group and organometallic species.
Sep 4, 2025 · Sawyer VanZanten, Corin C. Wagen
Which Optimizer Should You Use With NNPs?

Which Optimizer Should You Use With NNPs?

The results of optimizing 25 drug-like molecules with each combination of four optimizers (Sella, geomeTRIC, and ASE's implementations of FIRE and L-BFGS) and four NNPs (OrbMol, OMol25's eSEN Conserving Small, AIMNet2, and Egret-1) & GFN2-xTB.
Sep 4, 2025 · Ari Wagen and Corin Wagen
Double-Ended TS Search and the Invisible Work of Computer-Assisted Drug Design

Double-Ended TS Search and the Invisible Work of Computer-Assisted Drug Design

finding transition states; the freezing-string method; using Rowan to find cool transition states; discussing drug design
Sep 3, 2025 · Jonathon Vandezande, Ari Wagen, Spencer Schneider, and Corin Wagen
The Invisible Work of Computer-Assisted Drug Design

The Invisible Work of Computer-Assisted Drug Design

Everything that happens before the actual designing of drugs, and how Rowan tries to help.
Aug 28, 2025 · Corin Wagen
MSA, Protein–Ligand Binding Affinity Exploration, and Stereochemistry

MSA, Protein–Ligand Binding Affinity Exploration, and Stereochemistry

MSA-related occurrences and our incident postmortem; MSA server coming soon; exploring new approaches to binding-affinity prediction; a farewell to interns; a new stereochemistry lab
Aug 22, 2025 · Ari Wagen, Corin Wagen, and Spencer Schneider
Co-Folding Failures, Our Response, and Rowan-Hosted MSA

Co-Folding Failures, Our Response, and Rowan-Hosted MSA

A narrative account of our response to a sudden rise in protein–ligand co-folding failures.
Aug 22, 2025 · Ari Wagen
Exploring Protein–Ligand Binding-Affinity Prediction

Exploring Protein–Ligand Binding-Affinity Prediction

Trying a few modern ML-based approaches for predicting protein–ligand binding affinity.
Aug 20, 2025 · Ishaan Ganti