Variational AI is pleased to share that Rakovina Therapeutics (“Rakovina”) and Variational AI have expanded our collaboration to advance Rakovina’s kt-5000 series of ATR inhibitors, strengthening our joint commitment to accelerating oncology drug discovery through practical, design-first generative AI.
This expanded agreement builds on our existing work together and focuses on lead optimization of drug candidates generated by Variational AI’s Enki™ generative AI platform and selected by Rakovina for further evaluation in the kt-5000 ATR inhibitor program.
What’s new in the expanded collaboration
Under the expanded scope, Variational AI will apply Enki™ in an iterative optimization loop, generating and prioritizing multiple optimized compound designs to help Rakovina progress from early leads toward a clinical candidate faster, with the goal of achieving timelines measured in months rather than years.
As in our initial work together, Rakovina retains full control over which candidates to advance into laboratory testing and further development. Variational AI’s role is to support rapid design cycles that surface high-quality options; Rakovina determines which molecules move forward based on its strategy and experimental evaluation.
Importantly, this builds on the momentum already generated in the program. Rakovina has shared recent preclinical progress in the kt-5000 series, including results presented at the Society for NeuroOncology (SNO), highlighting efforts to advance CNS-penetrant ATR inhibitors. You can read a summary of those latest published results here: Update: Advancing CNS-Penetrant ATR Inhibitors with Enki.
What this means for our partners and the broader industry
This collaboration extension reflects a shared belief that the fastest path to high-quality candidates is not more screening, it’s better design cycles, supported by generative AI that can propose diverse, chemically realistic options while teams retain scientific and strategic control.
We’re excited to continue working with Rakovina as they advance the kt-5000 program and further explore the potential of ATR inhibition for patients who need better options.
Variational AI is excited to participate in the inaugural AI Convergence: Small Molecule Discovery Summit in Boston, bringing together leaders at the intersection of AI/ML, computational chemistry, and medicinal chemistry. This is a great venue to compare notes on how generative approaches are being applied to real-world small-molecule design workflows (from hit expansion to multiparameter lead optimization).
Our CEO, Handol Kim, will be speaking at the event.
Quality Over Quantity: Rethinking AI-Driven Drug Discovery Is Bigger Always Better? Training AI/ML Models with Low Volume Datasets to Generate Molecular Chemistry with Desired Preclinical Target Product Profiles & Physiochemical Properties
Handol will be available during the summit for conversations with pharma/biotech teams exploring AI-enabled small-molecule discovery. To schedule time, reach out to bd@variational.ai.
Building on the initial hit series Variational AI designed with Enki™, Rakovina Therapeutics presented new data at the AACR–NCI–EORTC 2025 Molecular Targets & Cancer Therapeutics meeting. The poster reports notable progress across potency, selectivity, metabolic stability, and in vivo PK/brain exposure for multiple Enki-generated ATR inhibitor chemotypes.
As a reminder of the workflow to date: Enki™ generated a prioritized list of 138 candidates against a preclinical TPP emphasizing ATR potency, selectivity against related PIKK family members, and CNS penetration; 35 were synthesized and advanced into wet-lab testing, delivering multiple hits with >50% showing ATR inhibition at 1 µM, in less than 12 months. Below is a summary of the new results.
Dose–response curves against recombinant ATR show that several Enki-designed compounds are as potent as—or more potent than—reference clinical-stage ATR inhibitors (including ceralasertib, tuvusertib, and elimusertib).
This confirms that Enki’s generative search can reach benchmark-level target potency while exploring novel chemical space. (Poster panel: “Compound Potency”.)
Selectivity across the PIKK family
Selectivity profiles indicate that our compounds match the PIKK-family selectivity of the clinical references, an essential requirement to minimize off-target liabilities typical of this kinase family. Maintaining this selectivity while pushing for CNS exposure is a key program goal, and the latest data indicate no trade-off to date. (Poster panel: “Selectivity vs. PIKK family kinases”.)
Metabolic stability
In human liver microsomes, compounds demonstrated favorable metabolic stability after 45 minutes of incubation, supporting suitability for in vivo studies and giving medicinal chemists a strong foundation for subsequent optimization cycles. (Poster panel: “Metabolic Stability”.)
Mouse pharmacokinetic profiling and CNS exposure
Following single-dose intraperitoneal administration (5 mg/kg) in mice, tested compounds were well tolerated and achieved measurable brain exposure, with varying CNS penetration across the series quantified by LC/MS in plasma and brain. This variability reveals structure–property relationships that we can exploit to systematically increase brain exposure without sacrificing potency or selectivity. (Poster panel: “Mouse Pharmacokinetic Profiling”.)
Why does this matter?
Achieving CNS penetration has limited many first-generation ATR programs, despite the compelling biology as a treatment approach for brain tumors and brain metastases. The new data suggest that CNS-accessible ATR inhibitors are attainable with Enki’s generative, multi-objective design, which jointly optimizes potency, selectivity, and ADME/CNS properties from the outset—rather than discovering potency first and back-filling other developability properties later.
Enki’s advantage stems from constructing a smooth chemical search space that supports efficient, multi-property optimization and rapid iteration with medicinal chemistry. This approach searches far beyond fixed screening libraries and prioritizes chemically valid, synthesizable designs tailored to the program’s TPP—driving both speed (design-to-hit in under a year) and quality (early selectivity, stability, and PK signals).
Conclusion & what’s next
These results illustrate what’s possible when generative AI augments traditional discovery:
Speed: from TPP to multiple validated hits and into in vivo PK/brain exposure studies in months, not years.
Quality: hits that already balance ATR potency, PIKK selectivity, and metabolic stability, with emerging CNS exposure that can be further optimized.
Next steps include continued in vitro / in vivo characterization and data-informed lead optimization to further elevate brain exposure while preserving selectivity and safety margins. The poster (embedded below) provides figure-level details and methods.
If you’re exploring targets where CNS penetration, multi-objective design, or fast hit-to-lead matters, we’d be happy to collaborate. Enki™ integrates smoothly with real-world medicinal chemistry workflows to deliver more—and better—shots on goal.
Collaboration aims to discover and develop novel small molecule therapeutics against two targets designated by Merck
VANCOUVER, Canada, [September 23, 2025] – Variational AI, Inc., a generative AI drug discovery company, today announced a collaboration with Merck, known as MSD outside of the United States and Canada, to apply Variational AI’s Enki™ platform to design and optimize novel small molecule candidates against two undisclosed targets.
Under the agreement, Variational AI will use a fine-tuned version of its Enki™ platform trained on Merck’s proprietary data to generate and optimize small molecule candidates against therapeutic targets designated by Merck. Merck will have the exclusive right to develop and commercialize compounds arising from the collaboration. Variational AI will receive an upfront payment and is eligible to receive milestones with a total potential deal value of up to USD$349 million.
“We are excited to apply our proprietary machine learning algorithms to Merck’s extensive and valuable datasets to create unique, fine-tuned generative models of unprecedented power and accuracy,” said Handol Kim, CEO of Variational AI. “This is a compelling framework that has the potential to significantly accelerate and redefine the unit economics of drug discovery.”
“At Merck, we are working to harness the potential of AI to improve efficiency, speed, and quality of candidates earlier in the discovery continuum,” said Robert M. Garbaccio, Ph.D., Vice President and Head of Discovery Chemistry, Merck Research Laboratories. “We look forward to working with Variational AI to apply their Enki™ platform to challenging therapeutic targets.”
Variational AI’s Enki™ platform is based on a foundation model trained on Variational AI’s internal data, a curated version of the totality of all publicly available data, and proprietary generative models to create and optimize small molecule leads based on partner target product profiles.
About Variational AI
Variational AI is a venture-backed generative AI drug discovery company based in Vancouver, Canada. Founded by machine learning researchers from MIT, Caltech, Google Research, Microsoft Research, and D-Wave Quantum, its proprietary Enki™ platform uses state-of-the-art generative AI models to design novel, optimized small molecules with improved probability of success, thereby accelerating the discovery and development of transformative therapeutics. For more information, please visit www.variational.ai.
Therapeutic resistance remains a major limitation in kinase inhibitor therapy. In non-small cell lung cancer (NSCLC)1, head and neck squamous cell carcinoma (HNSCC)2, and cholangiocarcinoma3, EGFR (epidermal growth factor receptor) is frequently overactivated and has been successfully targeted clinically with multiple generations of moderately selective small-molecule Tyrosine Kinase Inhibitors (TKIs)4 (e.g., erlotinib, gefitinib, afatinib, osimertinib). However, compensatory upregulation of FGFR1 (fibroblast growth factor receptor 1) and the development of tyrosine kinase mutations are well-documented resistance mechanisms in both preclinical and clinical settings5,6.
Studies have demonstrated that dual blockade of EGFR and FGFR1 can overcome acquired resistance and yield synergistic anticancer effects. However, despite this evidence, structurally validated dual EGFR/FGFR1 inhibitors are rarely reported in the literature and often carry additional kinase inhibitory activity. The chemical space for dual kinase inhibitors is not well-explored, and existing examples often involve non-selective or pan-kinase scaffolds.
A Shared Problem Sparks a Collaborative Approach
Following their successful 2022 collaboration, which targeted the SARS-CoV-2 main protease (Mpro), Variational AI (VAI) and Life Chemicals (LC) have partnered for a second time, this time focusing on kinase selectivity. Exploratory profiling of LC’s kinase-directed library against EGFR and aligned scientific interests led to a new joint program to discover de novo generated selective dual EGFR/FGFR1 inhibitors.
The strong biological and clinical rationale was faced with a technical challenge: public datasets contain an abundance of individual EGFR and FGFR1 inhibitor potency data. However, there were only 90 dual-active molecules (IC₅₀ < 300 nM) encompassing 7 hinge binding scaffolds. Many of these dual EGFR/FGFR1 active compounds had higher potency for other kinases and these activities were picked up in selectivity screens. To our knowledge, no compounds with selective dual EGFR and FGFR1 inhibition are under development. This dataset of dual active compounds represents a very small amount of data to train on for our desired activity profile. Additionally, dual kinase targeting poses significant design challenges given the structural similarity overlaps across the kinase family and limited chemical diversity. For example, EGFR and FGFR1 share only 44% sequence identity within their kinase domains7, which limits the extent of pharmacophoric overlap.
Despite these constraints, the two teams were excited to take on this challenge.
Tackling Multi-Target Design with Generative AI
To address the low-data challenge, the project deployed Enki™, VAI’s generative AI proprietary platform, to identify novel compounds. Enki™ is a fully-trained foundation model built around a variational autoencoder, a type of generative model that learns a smooth chemical latent space in which structurally novel molecules can be optimized for multiple properties simultaneously, such as potency, selectivity, ADMET, synthesizability, and more…8 It generates structures inherently optimized to meet a defined target product profile.
The project goal was to design, synthesize, and test approximately 20 compounds for the desired profile. The Target Product Profile used for Enki™ was optimized for:
Potency for both EGFR and FGFR19targets (predicted IC₅₀ < 100 nM)
Selectivity versus mitotic kinases: CDK1, AURKA (two kinases chosen to show initial representative selectivity)
Novelty in chemical structure (no known analogs or Bemis-Murcko scaffolds10)
Synthetic chemistry feasibility
A traditional method for generating hits would involve a high-throughput screen (HTS). However, using HTS to find dual active compounds is not typically done because the hit rates per individual target are already low and would be compounded when searching for dual activity without pan-kinase activity. Furthermore, HTS doesn’t give novel starting points unless one has access to a novel chemical library. As shown in this case study below, a single run of Enki for lead generation is approximately 5-10 times less expensive than a HTS on approximately 100,000 compounds and leads to novel lead generation for dual kinase active compounds with initial selectivity.
Output Shows Dual Potency and Novelty
From a single run, Enki™ was able to generate a collection of structures, in which:
40 diverse molecules that fit the Target Product Profile (TPP) were designed
21 diverse compounds were selected by Life Chemicals for route development
19 (90%) were successfully synthesized within the timeframe of the project.
Of the 19 compounds synthesized, the majority exhibited meaningful biochemical activity:
Active at either EGFR or FGFR1 kinase (IC₅₀ < 5 µM) in a biochemical assay11: 15/19 compounds
Dual EGFR + FGFR1 activity: 5/19 compounds
The Lead compound:
EGFR IC₅₀ = 29 nM
FGFR1 IC₅₀ = 175 nM
Selectivity:
100-fold over CDK1/cyclin B1
~10-fold over AURKA
Figure 1: The generic scaffold core and R-group vectors
Docking-based interpretation of selectivity:
FGFR1 & EGFR: Our top hit engages favorably with the hinge region in both kinases and fits well into the back hydrophobic pocket, supporting the strong observed activity.
Aurora Kinase A: While our lead compound maintains hydrogen bonding with the hinge, steric constraints in the binding site appear to reduce the compound’s ability to access the back hydrophobic interactions. The interaction pattern is shifted toward more solvent-exposed residues, likely contributing to reduced potency and explaining the ~10-fold selectivity.
CDK1: The compound fails to dock productively at the active site. The poor fit suggests a lack of stable binding interactions, consistent with the observed >100-fold selectivity.
Structural novelty:
All molecules were absent from SciFinder-n (April 2025)
Each featured a unique Bemis–Murcko scaffold
Tanimoto similarity for each compound was ≤ 0.5 vs. known EGFR and FGFR1 inhibitors
No prior dual EGFR/FGFR1 inhibitors shared the designed hinge-binding scaffold
Figure 2: The docking overlays of the blinded ligand in yellow to the EGFR and FGFR1 kinase binding site.
Conclusions and Path Forward
This collaborative effort demonstrates that generative AI can be used to design novel, synthesizable dual-active kinase inhibitors, even from a sparse training set. In just one generative cycle, multiple first-in-class scaffolds with dual EGFR/FGFR1 activity were identified, synthesized, and validated.
The current phase of the ongoing collaboration between Variational AI and Life Chemicals has focused on hit generation and chemical novelty. In the future, there is an opportunity to continue the development and validation of the most promising compounds using Enki™’s lead optimization capabilities.
The project remains active, and results continue to evolve, so stay tuned for further updates.
Interested parties are encouraged to reach out for discussion or potential collaboration.
Research conducted by the Variational AI team: Marawan Ahmad, John Boylan, Marshall Drew-Brook, Peter Guzzo, Ahmad Issa, Mehran Khodabandeh, Sara Omar, Jason Rolfe, and Ali Saberali.
And by the Life Chemicals team: Olga Balabon, Vasily Pinchuk, Alexander Shivanyuk, Taras Tymoshenko and Iryna Vashchenko.
We thank Amazon Web Services (AWS) for providing the cloud computing resources that made this research possible.
1) Chen G, Bao Y, Weng Q, Zhao Y, Lu X, Fu L, Chen L, Liu Z, Zhang X and Liang G (2020) Compound 15c, a Novel Dual Inhibitor of EGFRL858R/T790M and FGFR1, Efficiently Overcomes Epidermal Growth Factor Receptor-Tyrosine Kinase Inhibitor Resistance of Non-Small-Cell Lung Cancers. Front. Pharmacol. 10:1533. doi: 10.3389/fphar.2019.01533 ↩︎
2) Wang Z, Anderson KS. Therapeutic Targeting of FGFR Signaling in Head and Neck Cancer. Cancer J. 2022 Sep-Oct 01;28(5):354-362. doi: 10.1097/PPO.0000000000000615. PMID: 36165723; PMCID: PMC9523489. ↩︎
3) Qibiao Wu, Yuanli Zhen, Lei Shi, Phuong Vu, Patricia Greninger, Ramzi Adil, Joshua Merritt, Regina Egan, Meng-Ju Wu, Xunqin Yin, Cristina R. Ferrone, Vikram Deshpande, Islam Baiev, Christopher J. Pinto, Daniel E. McLoughlin, Charlotte S. Walmsley, James R. Stone, John D. Gordan, Andrew X. Zhu, Dejan Juric, Lipika Goyal, Cyril H. Benes, Nabeel Bardeesy; EGFR Inhibition Potentiates FGFR Inhibitor Therapy and Overcomes Resistance in FGFR2 Fusion–Positive Cholangiocarcinoma. Cancer Discov 1 May 2022; 12 (5): 1378–1395. https://doi.org/10.1158/2159-8290.CD-21-1168 ↩︎
4) Zubair T, Bandyopadhyay D. Small Molecule EGFR Inhibitors as Anti-Cancer Agents: Discovery, Mechanisms of Action, and Opportunities. Int J Mol Sci. 2023 Jan 31;24(3):2651. doi: 10.3390/ijms24032651. And, Liping Hu, Mengmeng Fan, Shengmin Shi, Xiaomeng Song, Fei Wang, Huan He, Baohui Qi, Dual target inhibitors based on EGFR: Promising anticancer agents for the treatment of cancers (2017-), European Journal of Medicinal Chemistry, Volume 227, 2022, 113963, ISSN 0223-5234, https://doi.org/10.1016/j.ejmech.2021.113963 ↩︎
5) Azuma K, Kawahara A, Sonoda K, Nakashima K, Tashiro K, Watari K, Izumi H, Kage M, Kuwano M, Ono M, Hoshino T. FGFR1 activation is an escape mechanism in human lung cancer cells resistant to afatinib, a pan-EGFR family kinase inhibitor. Oncotarget. 2014 Aug 15;5(15):5908-19. doi: 10.18632/oncotarget.1866. PMID: 25115383; PMCID: PMC4171601 ↩︎
6) Zhang, P., Yue, L., Leng, Q. et al. Targeting FGFR for cancer therapy. J Hematol Oncol 17, 39 (2024). https://doi.org/10.1186/s13045-024-01558-1; and, Tan L, Zhang J, Wang Y, Wang X, Wang Y, Zhang Z, Shuai W, Wang G, Chen J, Wang C, Ouyang L, Li W. Development of Dual Inhibitors Targeting Epidermal Growth Factor Receptor in Cancer Therapy. J Med Chem. 2022 Apr 14;65(7):5149-5183. doi: 10.1021/acs.jmedchem.1c01714. ↩︎
7) Modi, V., Dunbrack, R.L. A Structurally-Validated Multiple Sequence Alignment of 497 Human Protein Kinase Domains. Sci Rep 9, 19790 (2019). https://doi.org/10.1038/s41598-019-56499-4↩︎
8) Alperstein, Z., Cherkasov, A., & Rolfe, J. T. (2019, May 30). All SMILES Variational Autoencoder. arXiv.org. https://arxiv.org/abs/1905.13343 ↩︎
Vancouver, BC – July 31, 2025 — Variational AI is proud to announce its participation in a new open science drug discovery initiative led by Agora Open Science Trust. The project is focused on optimizing the potency and selectivity of lead inhibitors targeting PRMT6, a Type I protein arginine methyltransferase implicated in the pathogenesis of Spinal Bulbar Muscular Atrophy (SBMA), a rare neuromuscular disorder.
As part of the multi-institutional effort, the project aims to evolve the current PRMT6 chemical probe (SGC6870, IC₅₀ = 77 nM) into a preclinical lead compound. Variational AI will support the endeavor by leveraging its Enki™ generative AI platform to propose structurally diverse backup series for the project target product profile (TPP).
“This collaboration is a testament to Variational AI’s commitment to extending our Enki™ platform to tackling unmet medical needs — particularly in rare diseases with no approved treatments.” – Peter Guzzo, PhD, Executive Vice President, Drug Discovery at Variational AI.
“This partnership reflects a shift we are proud to lead, where open science is no longer seen as just an academic exercise. Contributions from industry partners, who provide access to cutting-edge technology like Variational AI, are critical to our mission to discover treatments for rare diseases like SBMA.” – Peter Sampson, Vice President of Drug Discovery, Agora Open Science Trust.
As two Canadian organizations, Variational AI and Agora are proud to advance a model of science that prioritizes transparency, collaboration, and global impact. The open science framework not only accelerates discovery but also ensures that promising medicines reach underserved populations more affordably.
About Agora Open Science Trust
Agora Open Science Trust is a Canadian charity whose mission is to accelerate the discovery and development of affordable new medicines through open science. Agora’s first initiative, M4K Pharma (‘Medicines for Kids’), is advancing a novel ALK2 inhibitor for the treatment of Diffuse Intrinsic Pontine Glioma. Agora’s pipeline of collaborative open science drug discovery programs has recently expanded to include Spinal Bulbar Muscular Atrophy, and Primary Sclerosing Cholangitis – both of which currently have no approved treatment. The SBMA/PRMT6 program is supported by funding from the Krembil Foundation and Conscience’s Developing Medicines through Open Science program. More information: https://www.agoraopensciencetrust.org
About Variational AI
Variational AI is redefining the unit economics of drug discovery for better patient outcomes. Founded in 2019, the company has built a proprietary generative AI foundation model purpose-built for small molecule discovery. Its Enki™ platform leverages this model to generate novel molecular structures de novo, optimized to meet a defined Target Product Profile. From Hit ID to Lead Optimization, Enki™ accelerates early discovery by generating synthesizable compounds without relying on high-throughput screening or traditional libraries. Variational AI is growing – check out our open positions to join the team! More information: https://variational.ai
VANCOUVER, BC – July 10, 2025 – Variational AI is proud to announce that it has been named Emerging Company of the Year – Biotech by Life Sciences BC. The recognition is part of the 27th Annual Life Sciences BC Awards, which celebrate outstanding achievements in British Columbia’s vibrant life sciences sector.
This award highlights Variational AI’s innovation in developing Enki, a generative AI foundation model that delivers novel and selective lead compounds to pharmaceutical and biotechnology companies. Variational AI’s unique approach leverages generative machine learning to efficiently explore the vast chemical space, bringing novelty in hit identification and accelerating lead optimization while delivering highly potent, selective, and synthesizable candidates.
“We are honoured to be recognized by Life Sciences BC as this year’s Emerging Biotech Company,” said Handol Kim, CEO of Variational AI. “This award is a testament to the remarkable work of our team and the potential of generative AI to transform how new therapies are discovered.”
The award will be formally presented during the Life Sciences BC Awards Gala on October 2, 2025, at the Vancouver Convention Centre West. The event will gather leaders and innovators from across the life sciences industry in British Columbia to celebrate innovation, excellence, and progress in life sciences.
For more information about the event and award recipients, please visit Life Sciences BC.
About Variational AI
Variational AI is redefining the unit economics of drug discovery through the power of generative AI. The founding machine learning (ML) team comes from leading AI research labs at Google, Microsoft, MIT, Caltech, and D-Wave Quantum working with established drug discovery leaders to develop Enki™, the industry-leading foundation model for small molecule drug discovery. Variational AI is based in Vancouver, BC, Canada, and is actively hiring. For more information visit us at https://variational.ai or email us at info@variational.ai.
Two AI-first biotech innovators join forces to co-develop novel small molecule therapeutics addressing chronic skin inflammation
Vancouver, BC – June 18, 2025 – Variational AI, developer of the Enki™ generative AI platform for drug discovery, today announced a co-development agreement with Oncocross (KRX:382150), a clinical-stage biotechnology company based in Seoul, South Korea, to co-develop a novel small molecule inhibitor for chronic skin inflammation, with a focus on atopic dermatitis (AD).
The collaboration combines the complementary strengths of both companies: Oncocross’s RAPTOR AI™ platform, which has identified novel insights to treat atopic dermatitis, as well as Oncocross’ proven clinical development capabilities, and Variational AI’s Enki™ drug discovery platform, which will be used to generate and optimize novel small molecules with high potency, selectivity, and developability.
“This partnership exemplifies the power of combining biological insights with generative chemistry to unlock new therapeutic possibilities,” said Handol Kim, CEO of Variational AI. “Oncocross’s disease-first AI expertise is a perfect match for our chemistry-first generative AI approach. Together, we are setting the foundation for a truly innovative drug discovery program.”
“This agreement reflects our belief that strategic and focused collaboration is the key to solving difficult medical challenges,” added Dr. Yi Rang Kim, CEO of Oncocross. “Different parts of the drug discovery and development process require different AI. Partnering with Variational AI will enable both of us to focus on what we do best, and to rapidly translate our disease models into promising small molecule drugs.”
About Variational AI
Variational AI is redefining the unit economics of drug discovery through the power of generative AI. The founding machine learning (ML) team comes from leading AI research labs at Google, Microsoft, MIT, Caltech, and D-Wave Quantum working with established drug discovery leaders to develop Enki™, the industry-leading foundation model for small molecule drug discovery. Variational AI is based in Vancouver, BC, Canada, and is actively hiring. For more information visit us at https://variational.ai or email us at info@variational.ai.
About Oncocross
Oncocross is an AI-powered drug development company using proprietary technologies like RAPTOR AI™, ONCO-RAPTOR AI™, and ONCOfind AI™ to drive therapeutic innovation across oncology, rare diseases, and inflammatory conditions. For more, visit www.oncocross.com.
How Variational AI supported Rakovina Therapeutics in advancing a novel oncology program
A Collaboration to Enable Faster Oncology Innovation
At the 2025 AACR Annual Meeting, Rakovina Therapeutics presented early-stage results from a program aimed at developing a brain-penetrant ATR inhibitor—an approach with potential relevance for patients with brain tumors or CNS metastases.
This initiative was supported by Variational AI’s Enki™ platform, which was used to rapidly generate and prioritize novel chemical structures aligned with Rakovina’s preclinical design objectives.
The Challenge: Limited Options for CNS-Targeting ATR Inhibitors
Ataxia telangiectasia and Rad3-related protein serine/threonine kinase (ATR) inhibition is a promising therapeutic strategy in oncology, because of its crucial role in DNA damage response. Yet many of the ATR inhibitors currently under development do not sufficiently penetrate the blood-brain barrier. This limits their application in indications involving the CNS.
Designing molecules that are both potent and selective for ATR—and possess the physicochemical properties required for CNS exposure—is a non-trivial challenge. It demands multi-parameter optimization and the ability to search widely and efficiently across chemical space.
The Role of Enki™ in Supporting the Program
Rakovina defined a clear target product profile (TPP) for a next-generation ATR inhibitor. Using this specification, Enki, our generative AI platform for small molecule design, was tasked to propose a diverse set of compounds that met the required criteria:
Potent and selective inhibition of ATR
Properties consistent with CNS penetration
Acceptable drug-like physicochemical properties for oral bioavailability
Enki generated and prioritized 138 novel compounds, which were then further refined into a shortlist of candidates for synthesis and in vitro testing. This front-loaded design effort enabled Rakovina to progress more quickly toward experimental validation.
From Molecule Generation to Real-World Testing
This collaboration underscores how AI can deliver more than incremental gains. By enabling rapid hit generation tailored to real-world design constraints, Enki empowers teams like Rakovina’s to move quickly from concept to candidate—without compromising on chemical novelty or developability.
The compounds generated by Enki are now undergoing in vitro and in vivo evaluation by Rakovina’s team. As noted in their public release, the ability to rapidly move from in silico candidate generation to preclinical testing highlights the efficiency gains made possible through AI-enabled design.
For Variational AI, this collaboration exemplifies how Enki can complement drug discovery teams by delivering high-quality leads tailored to difficult design challenges—particularly when traditional screening or structure-based methods may be limited.
But the implications go further. For any biotech or pharma team facing tight timelines, complex target product profiles, or tough-to-drug indications, Enki represents a new lever: faster, smarter chemical exploration—at scale.
Interested in Collaborating?
Whether you’re navigating lead optimization or building a new program from scratch, we’d love to explore how Enki can support your goals.
