Mutant selectivity remains one of drug discovery’s hardest problems. New preclinical research applying quantum chemistry to JAK2V617F illustrates how detailed molecular analysis can inform more selective inhibitor design.
Despite decades of research, achieving mutant selectivity remains a challenge in small-molecule drug discovery. Mutant-selective drugs are designed to inhibit disease-driving variants of a protein while sparing the wild-type form, but many targeted cancer therapies still act on both, narrowing therapeutic windows and leading to dose-limiting toxicities.
This challenge is well illustrated by JAK2V617F, a key driver mutation in myeloproliferative neoplasms (MPNs). Existing JAK inhibitors broadly suppress JAK signalling and are associated with cytopenias, a reduction in normal blood cell counts, highlighting the difficulty of selectively targeting the mutant protein.
Recent preclinical work from Prelude Therapeutics, supported by computational drug discovery company QDX, suggests that quantum chemistry–based simulations may offer a way to address this limitation. The collaboration led to the identification of a previously unrecognised binding pocket in the mutant protein, informing the design of JAK2V617F-selective inhibitors that spared wild-type signalling in in vivo models.
To explore how this work was carried out and what it could mean for early drug discovery, we spoke with Loong Wang, co-founder and CEO of QDX.
From high-performance computing to molecular design
Wang’s background lies outside traditional medicinal chemistry. He studied supercomputing at the Australian National University before founding a distributed computing company that later reached a market capitalisation of more than $1 billion. After selling the company, he moved into drug discovery, co-founding Automera, a company focused on autophagy-targeting degraders.
Our mission is to simulate the most complex systems nature has to offer, using the fastest quantum chemistry software ever built.
He now leads QDX, which applies large-scale quantum chemistry simulations to biological systems that have historically been modelled using classical computational approaches.
“Our mission is to simulate the most complex systems nature has to offer, using the fastest quantum chemistry software ever built,” Wang says.
The challenge of selectivity in JAK2-driven disease
JAK2V617F is present in a large proportion of patients with MPNs, a group of chronic blood cancers characterised by the overproduction of mature blood cells and plays a central role in disease pathogenesis.
“Current treatments broadly inhibit JAK2, so healthy cells are also affected,” Wang explains. “This leads to side effects and dose limitations.”
According to Wang, the key unmet need in this space is the ability to selectively inhibit the mutant protein while preserving normal JAK2 signalling. Achieving that level of discrimination has proven difficult using conventional design strategies.
Identifying a previously unknown structural feature
QDX was initially engaged to support structure-based optimisation of the JAK2V617F inhibitor programme using its quantum chemistry platform. During the course of the work, early signals of selectivity began to appear, but without a clear mechanistic explanation.
Our simulations were able to discover that a previously unknown pocket was being opened by some inhibitors and only in the JAK2V617F.
“During the programme, as we were helping to optimise various properties, some selectivity started being noticed,” Wang says.
Using a combination of quantum mechanics and molecular mechanics (QM/MM) simulations, QDX explored the system in greater detail.
“Our simulations were able to discover that a previously unknown pocket was being opened by some inhibitors and only in the JAK2V617F.”
Wang says the finding helped explain the selectivity being observed and informed subsequent compound design.
Why quantum chemistry was required
Wang argues that identifying this pocket would not have been possible using classical computational approaches.
“The main advantage of QM simulations is that they accurately capture all physical effects with far fewer approximations and assumptions than classical approaches. The main disadvantage is that QM has historically been too slow and expensive to be used at any practical scale,” he says.
QM was needed to get sufficient predictivity for filtering and designing.
QDX’s platform is designed to address this limitation, enabling large-scale simulations that predict binding poses, binding affinities and, in some cases, properties relevant to absorption, distribution, metabolism and excretion.
The team also trained generative AI models on QM-derived data to propose molecular modifications during optimisation.
“We did try classical techniques, but they performed incredibly poorly and couldn’t produce anything except noise,” Wang says. “QM was needed to get sufficient predictivity for filtering and designing.”
Preclinical evidence of mutant selectivity
Prelude reported that the resulting compounds selectively targeted JAK2V617F-positive stem and progenitor cells while sparing wild-type cells. In mouse models, Prelude reported that treatment normalised blood counts and spleen size without inducing the cytopenias typically associated with JAK inhibition.
“This suggests that these compounds can distinguish malignant cells from healthy blood‑forming cells, raising the possibility of deeper reduction of the disease‑driving clone with a more favourable safety profile. Of course, this will need to be confirmed in human trials.”
Implications for early drug discovery
Beyond this specific programme, Wang sees broader implications for how computational chemistry is used in early drug discovery. He argues that accurate, fast simulations can substantially reduce experimental burden by filtering designs before synthesis.
In short, it is far more efficient to run a calculation than it is to synthesise a design and run an experiment.
“In short, it is far more efficient to run a calculation than it is to synthesise a design and run an experiment,” Wang says. “If your calculations are accurate and fast, you can use them to filter designs and not bother making the ones that are predicted to perform poorly.”
He also points to the ability of these approaches to reveal mechanisms that may be difficult or impossible to observe experimentally.
“Super accurate simulations can offer insights that experiments can’t because you can pause, fast forward, zoom in, zoom out, rewind and replay,” Wang says. “You can see how every single electron moves with time and, at least right now, that’s something no experiment will give you.”
Next steps
While Wang does not go into detail on the next steps for the current inhibitors, he says the discovery has broader significance for the field.
“Our discovery opens the door for a whole new type of selective JAK2V617F inhibitors, targeting a novel mechanism for not just isoform selectivity, but mutant selectivity. We now have a new way to achieve this that wasn’t previously known and therefore hasn’t been fully exploited yet.”
He expects future programmes to build on this approach, using quantum chemistry to uncover subtle structural differences that enable selective targeting of disease-driving variants.
As quantum-level simulations become more scalable, Wang suggests they could play an increasingly central role in early drug discovery, particularly where traditional approaches have struggled to deliver selectivity.
About the expert
Loong Wang is co-founder and CEO of QDX, a computational drug discovery company applying large-scale quantum chemistry to biological systems. He studied supercomputing at the Australian National University and has founded multiple technology companies across computing and biotechnology. Prior to QDX, he co-founded Automera, an autophagy-focused drug discovery company and previously built and exited a distributed computing business. His work focuses on using quantum-level simulations to support early-stage drug design.
Related topics
Artificial Intelligence, Assays, Cancer research, Computational techniques, Drug Discovery, Drug Discovery Processes, Drug Targets, Medicinal Chemistry, Molecular Modelling, Oncology, Precision Medicine, Small Molecules, Structural Biology, Translational Science