By Zhen Chen, Senior Principal Scientist

Solving the biggest challenge in RNA-targeted DEL screens with a mirror-image approach.

RNA-targeted small molecule drugs are one of the most promising frontiers in drug development, but they’re still surprisingly underexplored. That’s starting to change. A major milestone came in 2020 with the FDA approval of Genentech’s Risdiplam, an RNA splicing modulator that has significantly improved outcomes for people living  with spinal muscular atrophy (SMA). Since then, interest in this space has only grown. Companies like Arrakis, Ribometrix, Lilly and others are actively building pipelines for RNA modulators.

That said, finding small molecules that effectively target RNA isn’t easy. Current methods such as affinity selection-mass spectrometry (ASMS) and phenotypic screening have their strengths, but they’re often limited in scale and throughput. That’s where DNA-encoded library (DEL) screening comes in. DEL can accelerate progress by exploring much larger chemical space, giving drug discovery efforts a serious boost right from the start.

However, historically, the perception is that RNA is not amenable to DEL screening. This is because the DNA tags on DEL can hybridize with the RNA target itself, resulting in spurious enrichment of false positives. Earlier studies from WuXi and HitGen have introduced various blocking methods to minimize this kind of interference. But even when hybridization is reduced, it can still produce false hits and bury the enrichment signal of real binders. In response to growing interest in RNA-targeted drug discovery, we’ve established a patent-pending strategy, “mirror image DEL screen,” a major step forward in tackling the challenges of this space.

X-Chem’s Mirror-image DEL screen compared to conventional DEL screen for RNA targets

Inspired by Sczepanski and Joyces’ cross-chiral SELEX method, we perform an affinity-mediated screen with our collection of DEL against an L-RNA target, which is the mirror image of natural D-RNA. This chiral form preserves the sequence and folding conformation of natural D-RNA but does not hybridize with DNA tags, helping us avoid the false positives that have historically muddled with true binding signals. The mirror-image versions of the L-RNA-enriched hits are then deduced to be binders of the natural D-RNA and can be confirmed as genuine RNA target engagers by off-DNA synthesis and orthogonal assays.

How X-Chem’s mirror-image DEL screen for RNA targets work

How does our new approach affect DEL screening outcome? The difference is striking. When we screened our DEL against mirror-image L-RNA targets, we saw no evidence of nucleotide sequence-based enrichment. This elimination of DNA:RNA hybridization translates to the complete removal of false hits. For comparison, under similar conditions using natural D-RNA, DNA sequence motifs were enriched up to 1000-fold.

In addition, we consistently observed >10-fold enhancement in enrichment signal from real binders (both spiked-in positive controls and de novo discovered hits) toward L-RNA compared to natural D-RNA. For difficult targets like RNA, enrichment enhancements of this magnitude can often uncover new hits and turn failed screens into productive ones.

For further proof of concept, we screened our DEL deck of more than 100 billion compounds against three distinct RNA targets – an expansion repeat, a splice site, and a riboswitch. In each case, we discovered novel chemical matter with favorable physicochemical properties. After synthesizing the mirror-image versions of the DEL hits, we confirmed their binding to their respective natural RNA targets in orthogonal biophysical assays like SPR and/or ASMS. Further details can be found in our scientific poster here.

By eliminating DNA:RNA hybridization, our mirror-image strategy addresses the biggest barrier in RNA-directed DEL screens. RNA is now just as amenable to DEL screens as any other target class. With this technology, our partners can confidently screen RNA targets against more than 100 billion diverse drug-like compounds – 6 orders of magnitude higher than other screening methods.

And we don’t stop at hit discovery. Every confirmed hit comes with rich structure-activity relationship (SAR) data, which our team can help you build pharmacophore models and machine learning tools to guide optimization and accelerate progress toward clinical candidates.

In short, our DEL platform is now fully equipped to realize the potential of RNA-targeted drug discovery. We’re excited to help you make the most of it.

By Matt Clark, Chief Scientific Officer

One of the best parts of being part of X-Chem is seeing all the great science being conducted around the company every day. Our clients know that we always do our utmost to solve their problems and create scientific value. But little of that work is ever shared with the outside world. This fact is not unique to X-Chem; the entire biopharma industry is slow to publish their research, for the obvious reasons that it can help enable competitors and erode competitive advantage. Fortunately, when you’ve been around as long as X-Chem has, just enough material trickles into the public domain to give the community a glimpse of our scientific developments.

In that light, 2024 has been a breakthrough year for X-Chem’s innovations. We had 12 peer-reviewed publications this year, and additional posters and conference talks around the world. These disclosures underscore the collaborative potential of our drug discovery approaches.

Our team has advanced synthetic chemistry methodologies in several key areas. We reported novel approaches including deuteration through flow chemistry, amination of heterocycles using desaturative catalysis, and synthesis of novel amino acid building blocks via Negishi coupling. Our goals remain consistent: expanding the chemists’ toolkits to create compounds with potential bioactivity to improve human health.

As part of this broader effort, we are also closely following emerging trends in the field of covalent drug discovery, which has become increasingly important in targeting challenging proteins. We conducted a patent landscape analysis that focused on selective covalent reacting groups used in FDA-approved drugs. The analysis identified over 700 patent applications from more than 300 organizations, revealing strong interest in targeting key proteins like BTK, EGFR, KRAS G12C, and SARS-CoV-2 M^pro. This surge in innovation around targeted covalent inhibitors aligns with X-Chem’s own efforts to develop covalent modulators using our technology platform. As covalent drug discovery continues to evolve, we are excited to contribute our expertise to this rapidly developing field, helping to push the boundaries of what is possible in drug discovery.

A cornerstone of our research has been DEL screening, which we’ve shown through diverse collaborative projects. One example was our work with Texas A&M, from which the hits directly from DEL screen demonstrated in vivo efficacy in a mouse model for tuberculosis. In another significant collaboration with AstraZeneca, we developed a novel approach based on the cyanoacrylamide motif for discovering reversible covalent inhibitors of the oncology target Bfl-1.

Our contributions to the scientific community through Perspectives highlights X-Chem’s thought leadership in the evolving field of DEL and its growing role in drug discovery. One of these Perspectives delves into the application of DEL in covalent ligand discovery, a field that has advanced significantly since we first reported on covalent DEL at the DEL symposium in 2015. Another Perspective, written in collaboration with AstraZeneca, examines the structural biology of DEL hits reported in the literature. This paper demonstrates that DEL screening tends to yield hits with novel binding modes and unexpected interactions with targets, in most cases resulting in exquisite selectivity. We often talk about the chemical diversity of DELs with respect to other libraries, but this paper shows that the true measure of chemical diversity is the wide range of biological diversity that can be found in the library. We see so many unprecedented interactions from DEL because we are surveying so much unprecedented chemical space.

Our work goes beyond hit identification, with several examples illustrating the progression from optimizing DEL hits into lead- and candidate-quality chemical matter. With Symeres, we described the discovery and optimization of macrocycles targeting the oncology target Mcl-1. While large and peptidic, these compounds showed acceptable permeability in addition to high potency, presumably aided by the observed transannular hydrogen bonding.

Similarly, our work with Abbvie on IL-17 inhibitors showed the power of DEL-generated data to provide structural insights. Aided by the DEL-data generated SAR and co-crystal structural information, the Abbvie team executed a very impressive medicinal chemistry campaign to arrive at a compound with good PK properties and potent efficacy in an animal model.

One of the most impactful projects was our collaboration with Arbutus on the development of inhibitors for SARS-CoV-2 M^pro. In a field where nirmatrelvir, a covalent M^pro inhibitor, was the first approved therapy, we identified a novel noncovalent inhibitor from DEL screening. The challenge with current Mpro inhibitors is that they often have limited efficacy against nirmatrelvir-resistant variants and may not offer broad-spectrum activity against emerging coronavirus strains, which poses a significant risk in future outbreaks. In our approach, we first used DEL screening to identify a promising noncovalent Mpro inhibitor with a distinct binding mode compared to nirmatrelvir. Our chemists further enhanced this lead by employing a macrocyclization strategy to lock the active conformation, leading to a remarkable 10-fold potency improvement.

In addition to target-specific generation of chemical matter, we explored ligand generation for a broad target class in collaboration with Structural Genomics Consortium (SGC), Google, and Relay Therapeutics. By conducting large-scale multiplexed DEL selection followed by machine learning (ML) model training, active compounds were identified from commercial sources for 7 out of 16 screened targets, all of which were previously unliganded! This paper provides yet another proof point that high-quality DEL data is critical to drive the success of ML in drug discovery.

X-Chem helps partners advance by producing key discoveries that they can build upon. In a recent publication, our SGC colleagues used a DCAF1 ligand that we had identified through DEL screening to generate a functional PROTAC and fully characterize the structure of its ternary complex. This is the first example of a DEL-derived E3 binder supporting targeted protein degradation. And notably, DCAF1 was previously unliganded until our discovery in 2023.

To sum up, the state of scientific excellence at X-Chem is strong. Our approach to discovery is deeply rooted in collaboration. Our ultimate vision is to help create a healthier world by enabling others to achieve groundbreaking innovations. By continually enhancing capabilities in areas such as computational chemistry, DMPK, and data sciences, we are well-positioned to support our partners and clients in advancing their drug discovery efforts. Here’s to a great 2024, and even greater achievements with our partners in 2025!

By Ali Chou, Research Scientist

When it comes to drug discovery and development, having the right team and partners is critical for success. For a small biotech, good CROs can be force multipliers, bringing deep and complementary expertise in key areas of drug discovery. The partnership between X-Chem and Montara Therapeutics, led by founder and CEO Dr. Nicholas Hertz, is an example of an innovative, client-first collaboration with a joint goal of finding therapeutics for patients with neurological diseases.

Before founding Montara Therapeutics, Hertz co-founded and served as the CSO of Mitokinin, a biotechnology company that developed PINK1-targeted therapeutics to treat neurodegenerative diseases. His team had worked closely with X-Chem to discover and optimize compounds targeting PINK1. Their work was successful, leading to the development of a patented clinical candidate and the acquisition of Mitokinin by AbbVie in 2023.

Shortly after, Hertz co-founded Montara Therapeutics. Leveraging a brain-specific pharmacology platform published by Kevan Shokat’s group, they aim to create next-generation therapies for neurological diseases, including dementia. For this new venture, Hertz renewed his collaboration with X-Chem based on their previous experience and X-Chem’s expertise in brain-penetrating drug candidates.

“What really sets X-Chem’s team apart from other CROs is their scientific due diligence and commitment to innovation,” said Hertz.

According to Hertz, in 2022, during Mitokinin’s work with X-Chem, a paper from David MacMillan’s group appeared in Science. The X-Chem team contacted Hertz three days after the publication.

“The paper was about a new way to install different types of bonds. X-Chem reviewed the publication just a few days after being published and proposed to apply this method to our project because it could help us access unprecedented chemical space,” said Hertz.

For drug hunters, accessing completely new chemical space means a path to differentiated pharmacology and patentable molecules.

“Anyone can do basic molecule optimization,” said Hertz, but he noted that X-Chem’s proactive approach to bringing innovation and value to the project really made a deep impression on him.

When Mitokinin was founded, Hertz and his team evaluated all possible CROs in Western countries with extensive neuroscience experience.

“X-Chem had many publications in this area. They specialized in optimizing molecules for brain penetration and efficacies in the central nervous system.”

A handful of companies made it to the shortlist, but according to Hertz, what stood out was that X-Chem’s team really understood how to integrate its chemistry expertise with Mitokinin’s biological discoveries.

“They’re a very cohesive and dedicated team. They do a really good job,” said Hertz about the X-Chem team. “The biggest testament to the excellence of your team is if people come back to work with you again.”

The right CRO with specialized expertise can be tremendously impactful for the success of a drug discovery program. Top CROs bring not only robust capabilities but also a solutions- and innovations-oriented mindset to push projects forward, even when dealing with significant challenges.

For companies working in therapeutic areas with huge unmet patient needs, having a CRO partner with a track record of innovation and a deep commitment to science is invaluable. As Montara continues to develop new classes of neurological therapies, X-Chem is an indispensable partner for pushing the boundaries to maximize its client’s success.

By Ferenc Andreáni, Director of Chemistry

Drug Hunter recently wrote the article “Decoding DNA-Encoded Libraries for Drug Discovery.” Over the past two decades, DEL technology has evolved to a broadly used hit identification platform in drug discovery. Compared to other approaches like high-throughput screening (HTS), which on average costs $0.4-2 billion to generate a 1-million compound screening collection, an 800-million compound DEL screen library costs less than $1 million. The value of DEL screening speaks for itself.

A robust screen not only relies on the design and execution of the selection experiments (which we discussed in this webinar) but also the quality and diversity of the building blocks (BBs) used to construct DEL. Because the reaction conditions during DEL synthesis must be compatible with DNA, it poses limitations on the types of BBs that could be used in making DEL. Although several vendors (e.g., Enamine, Millipore Sigma) manufacture BBs, there is still a perceived shortage of diverse and novel BBs on the market.

In our previous blog, we described how we leverage AI to design and validate BBs. We use an iterative active learning loop consisting of experimental validation in the lab to generate data and ML models trained on the data to predict the validation outcomes of a new set of BBs. Then, we experimentally validate the predictions and optimize the ML models. The outcome: saving weeks of time for library development.

With AI facilitating the validation of BB reactivity, we’ve synthesized more than 1,000 novel BBs, 700 of which are available to purchase. We show a fraction below while more information is here.

All BBs are synthesized with these considerations:

• Low aromaticity and high fsp3 character which correlates with increased three-dimensionality, allowing more complex 3D structures and better target specificity; in addition, most compounds with high fsp3 characters present improved solubility, lipophilicity profiles, higher developability and enhanced metabolic stability
• Extended chemical space as shown below, where our designs (right panel, pink dots) cover new chemical space compared to existing XBBs on the market (left panel, grey dots). We also include bi- and tri-functional unnatural amino acids (heterocycles condensed with or connected to hetaryl/spiro rings via C-C, C-N bonds or C-O bonds), halo-acids, protected amino acids and halo-amino-acids

• Novelty and innovation are guaranteed as we check against eMolecules and our internal database to ensure that all BBs designed have not been synthesized before

The integration of AI-driven approaches in building block design, coupled with the advancements in DEL technology, marks a significant leap forward in drug discovery. By addressing the critical need for diverse and novel building blocks, companies like X-Chem are paving the way for more efficient and effective hit identification. As this technology continues to evolve, we can anticipate a broader range of drug candidates, faster discovery time lines and potentially lower costs in the pharmaceutical development process. The future of drug discovery looks promising, with DEL technology and AI-designed building blocks playing pivotal roles in unlocking new therapeutic possibilities and potentially bringing life-changing treatments to patients more rapidly than ever before. Contact us today to discuss your BB needs.

By Mounir Raji, Senior Scientist

Isotope-labeled compounds have emerged as important tools in drug discovery and development. The incorporated isotopes, particularly deuterium, in pharmaceuticals, have been shown to improve ADME properties compared to their non-isotopically labeled counterparts. Since 2006, the FDA has approved >10 drugs labeled with stable isotopes such as deutetrabenazine (deuterated), deucravacitinib (deuterated) and lubiprostone (C13-labeled).

Traditionally, isotope-labeled compounds are generated by different methods, e.g., organic chemistry routes in batches using isotopically enriched starting materials. At X-Chem, we took two synthetic directions to improve the cost-effectiveness, regioselectivity, versatility and the yield of isotope-labeling reactions. We specifically developed two methodologies to address current challenges in deuterium- and C13-labeling:

1. Using late-stage direct hydrogen isotopic exchange (HIE) instead of isotopically enriched starting materials: Direct isotopic exchange allows us to perform the labeling without the need for prefunctionalization. It’s also cost-efficient, due to the availability of reagents, such as D₂O or ¹³CO₂. Atom-economy-wise, the labeling can be introduced at a specific position without constructing the entire molecule, which adds to another advantage — more applicable to synthesizing complex natural products or pharmaceuticals.

2. Using flow synthesis instead of batch synthesis: Flow chemistry can offer numerous advantages compared to traditional batch chemistry, such as precise control of reactions parameters (e.g., temperature, reaction time), improved mixing and enhanced safety. We surmised that isotope labeling can also benefit from flow conditions.

Highly selective deuterium labeling

Inspired by Sajiki’s work using Pd/C, D₂O and catalytic amount of D₂ gas to label nucleobases, we screened different substrates with several readily available heterogenous catalysts under flow conditions. Of all the tests, Raney nickel had shown superior deuterium incorporation and could be fine-tuned by pretreatment with caffeine. From those results, we then optimized the reactions to develop a continuous flow Raney nickel-catalyzed HIE process, which was published in Advanced Synthesis & Catalysis. The process is compatible with nitrogen-containing heterocycles and pharmaceutical compounds, as we successfully labeled various purine bases, imidazole, pyridine as well as complex pharmaceuticals such as abacavir and remdesivir at high yields.

Controlled ¹³CO₂ gas generation

The most common source of ¹³C is ¹³CO₂ gas, which is typically available in cylinders. Achieving sufficient mixing and controlling the stoichiometry of multiphase (e.g., gas-liquid) reactions can be challenging. Moreover, these cylinders often necessitate excess pressure and are costly.

As a solution, we developed an on-demand ¹³CO₂ flow chamber.

The chamber is based on a “tube-in-tube” flow reactor. This gas-liquid reactor comprises of a Teflon membrane encased in a PTFE tube. This design leverages the principle of diffusion to facilitate the transfer of the reactive gas into the solution stream during the synthesis. The use of Na₂¹³CO₃ as a precursor allows for greater control over reaction stoichiometry, yielding a series of C13-labeled products such as bioactive ureas and ¹³C-carboxylic acids with high efficiency.

Applications Beyond Pharmaceuticals 

Deuterated and C13-labeled compounds are used in various fields even beyond pharmaceuticals or metabolic studies. For example, in mass spectrometry, labeled compounds serve as internal standards for quantification. In kinetic isotope effect studies, labeled compounds are used to interrogate reaction mechanisms and transition state structures. We envision that with our improved synthetic methods, isotopically labeled compounds could be generated more efficiently and cost-effectively.

By Dylan Hale, Research Scientist, and Arnaud Clerc, Research Scientist

When chemists approach the formation of a new bond, they often choose reaction conditions that they (or their colleagues) are familiar with and have worked for them in the past. Typically, they would first run a test reaction on a small scale (5 mg), evaluate the outcome and make changes to the reaction conditions that they believe will give a more favorable result. This iterative process is then repeated ad infinitum until they are satisfied with the result, and the desired product is isolated. But what happens when the typical conditions give little to no product? Or there is little precedent for the desired transformation? Or the number of variables that require optimization is simply too high?

The development of high throughput experimentation (HTE) has been a transformative technique in the pharmaceutical industry as it allows for the investigation of many variables at the same time. At X-Chem, we’ve developed a specialized team for high throughput route optimization (HTRO) that works alongside other chemists to unlock synthetic roadblocks by using HTE.

Figure 1. Workflow of HTRO at X-Chem.

The HTRO team at X-Chem leverages expertise in organometallics, catalysis, method development and design of experiments to uniquely tailor screens to every chemist’s needs. In addition to HTRO for photoredox, hydrogenation and reductive aminations, transition metal-catalyzed cross-coupling reactions are by far the most requested transformation for our team. Thus, we have established a streamlined library of catalysts and ligands that we can rapidly deploy in fit-for-purpose screens. We design these screens based on literature precedent, our chemist’s knowledge and our own experience. We strive to include outside-the-box examples in our screens to provide opportunities for engineered serendipity.

To achieve our throughput while remaining conscious of material constraints, we combine ChemBeads technology (developed by AbbVie) with a “pool and split” approach to our experimental design (introduced by BI). ChemBeads, a form of solid dilution, allows us to accurately dispense sub-milligram amounts of material, which in turn minimizes the material investment by the chemist requiring HTRO. Typically, a full 96 well plate requires less than 20 mg of material. Reaction conditions are rapidly identified by applying a combinatorial “pool and split” approach to the design of our screening platforms. By combining metals, ligands and other reactants into so called “pools” based on their chemical properties, we can survey hundreds and even thousands of unique conditions in 96 individual reactions. After identifying the pools that provided the highest amount of product, we identify the right conditions by deconvoluting the pools into their individual components. We found that this approach provides conditions robust, scalable and reproducible by the chemists.

When it comes to maximizing impact and minimizing materials, the HTRO team has saved months of chemists’ time and grams of advanced intermediates. Our ability to cover the breadth of chemical space with the least amount of material has been an enabling tool for the X-Chem team, and we continue to grow our capacity every day.

By Stéphane Turcotte, Senior Principal Scientist

If the bond is forever, it needs to be safe.

Covalent drugs have proved to be successful therapies for various indications. However, due to safety concerns, covalent compounds are not always considered when initiating a target-directed drug discovery project. Methods that can accurately assess the reactivity of irreversible and reversible covalent warheads at the start of a project would be valuable in derisking downstream drug development. We recently published a perspective in ACS Chem Biol on current approaches in covalent DEL technology and future steps to advance the field.

In a drug discovery project, to ensure that potent covalent inhibitors are not promiscuous, electrophilic compounds are tested for reactivity with thiols. This measurement is carried out using a glutathione (GSH) reactivity assay, where a compound is incubated with excess GSH, and the consumption of the compound is measured by liquid chromatography-mass spectrometry to determine a pseudo-first order rate constant.¹ A GSH half-life can then be determined from this rate constant. Safer drugs typically have a longer GSH half-life.

The current industry standard GSH reactivity assay has several limitations. It treats irreversible and reversible covalent warheads in the same way. It is concentration-dependent, it is difficult to reproduce the same results across different labs and it is not always predictive of toxicity. At X-Chem, we developed a new GSH assay to address some of these limitations (Fig. 1).

Figure 1. (A) The industry standard GSH reactivity assay is used to measure reactivity through GSH half-life, compared to (B) X-Chem’s GSH reactivity assay for reversible covalent warheads, which measures GSH k_off rather than reactivity, and (C) X-Chem’s GSH reactivity assay for irreversible covalent warheads, which is used to compare the GSH reactivity relative to the protein target with the same second order rate constant, k_inact/K_I.

Our new GSH assay is like the industry standard in terms of the procedures involved in wet work, and is also high throughput, but we take additional parameters into consideration when analyzing the data, so that we can solve for the kinetics of warhead reactivity.

For example, for reversible covalent warheads, we use off-rate and not reactivity rate as a predictor for toxicity. We measure the equilibrium concentration of GSH-adduct at different GSH concentrations, then we determine K_d and off-rate (k_off) for the reversible reaction. We assess toxicity by the difference between residence time (1/k_off) on the protein target and on GSH.

For irreversible covalent warheads, we measure the increase of GSH-adduct concentration with time at different GSH concentrations. We then plot the different observed rate constants together to solve for k_inact and K_I. This method of assessing reactivity is akin to the FDA’s requirement for CYP inhibition and is also the method that covalent drug researchers are already using to assess on-target reactivity. We assess toxicity by the difference between k_inact/K_I for the protein target and GSH.

By changing how we extract kinetic constants from reactivity data, we can generate predictors for toxicity and differentiate the kinetics for reversible and irreversible warheads. Moreover, our readouts are in line with the best practice for adduct analyses of covalent proteins. This assay not only helps us design and incorporate more refined covalent building blocks (the structures of which are based on approved and late-stage drugs that target Cys, Lys, Ser) for our DNA-encoded libraries for screening, but also provides our partners with better tools to optimize and derisk their covalent hits and leads.

Reference

1. Boike, L. et al. Advances in covalent drug discovery. Nat Rev Drug Discov., 2022, 21, 881–898.

Additional Reading:

1. Gai, C. et al. Advanced approaches of developing targeted covalent drugs. RSC medicinal chemistry, 2022, 13(12), 1460-1475.
2. Faridoon, Ng R. et al. An update on the discovery and development of reversible covalent inhibitors. Medicinal Chemistry Research, 2023, 32(6), 1039-1062.
3. Strelow, J.M. A Perspective on the Kinetics of Covalent and Irreversible Inhibition. SLAS Discovery, 2017, 22(1), 3–20.
4. Brum, L. In Vitro Drug Interaction Studies – Cytochrome P450 Enzyme and Transporter-Mediated Drug Interactions Guidance for Industry.

By Johan Bartholomeus, Senior Principal Scientist, and Philippe McGee, Principal Scientist

In a drug discovery project, during the hit-to-lead and lead optimization phases, pharmacokinetic parameters (absorption, distribution, metabolism, excretion [ADME] and toxicity) become increasingly important because they can indicate which drug candidates have the greatest therapeutic potential with the lowest risks. A commonly overlooked technique called K_IAM is of particular interest to X-Chem, as this technique can predict aspects of absorption, target promiscuity and certain types of toxicity — think a stitch in time saves nine (Fig. 1).

Figure 1. LogK_IAM is more therapeutically relevant than logD.

Log K_IAM refers to the partition coefficient (K) of a compound between an aqueous phase and an immobilized artificial membrane (IAM). This value can be used to predict a compound’s likelihood of passive diffusion and, hence, the compound’s potential for passage across biological membranes.

What K_IAM can also indicate is the probability of a drug candidate to display off-target promiscuity or induce phospholipidosis, a condition where excess phospholipids and drugs accumulate inside lysosomes. Under normal conditions, lysosomal phospholipase A2 (LPLA2) is bound to the inner membrane of the lysosome through electrostatic charges. There, phospholipase digests phospholipids, which is a natural process of metabolism.

However, in situations where phospholipidosis is induced by a drug, the drug accumulates at the inner membrane of the lysosome. This eliminates the anionic binding sites for the phospholipase, resulting in its degradation and toxic accumulation of phospholipids. Compounds with log K_IAM values > 4 can carry a phospholipidosis risk and should be considered carefully before progressing.

Although log K_IAM is just one factor to consider when assessing the potential risk of target promiscuity or phospholipidosis associated with a compound, we at X-Chem think that it is good practice to deprioritize compounds with high log K_IAM values to derisk downstream drug development efforts.

Log K_IAM values can be obtained simply through HPLC (Fig. 2). The analyte is introduced in the chromatographic system where it interacts with the stationary phase — the phospholipid IAM and the mobile phase — the aqueous buffer. The retention time of the analyte is calibrated to a set of reference compounds to calculate the partition coefficient, which is then converted to log K_IAM.

Figure 2. IAM HPLC column mimics the surface of biological cell membrane and can be used to predict passive diffusion of the drug.

The log K_IAM value can be considered a more biologically relevant variant of log D. We systematically measure log K_IAM for all project compounds synthesized at X-Chem as a routine part of the QC process.

In conjunction with other measured parameters, log K_IAM values can help provide early-stage predictions of the properties and behavior of synthesized compounds in vivo.¹ By studying the retention time, peak shape and elution profile of a drug on a biomimetic column, we can improve our ADME and toxicity predictions.²

For example, we observed a clear relationship between log K_IAM and the fraction unbound in microsomes (Fig. 3). Since compound stability in microsomes can increase with increasing proportions bound to microsomal macromolecules (as such, protecting the drug from being metabolized), rank ordering compounds by their total clearances risks reflecting differences in microsomal binding rather than actual metabolic stability.

Figure 3. Relationship between log K_IAM and the fraction unbound in microsomes.

We believe that log K_IAM will be a standard measurement in the future the way that logD is now. Our clients will benefit from getting K_IAM values on all project compounds. Log K_IAM data will allow us to advance projects more quickly, creating compounds with the right balance of properties to eventually become development candidates.

References

1. Casartelli, A. et al. A cell-based approach for the early assessment of the phospholipidogenic potential in pharmaceutical research and drug development. Cell Biology and Toxicology, 2003, 19(3), 161-176.
2. Valko, K.L. Application of biomimetic HPLC to estimate in vivo behavior of early drug discovery compounds. Future Drug Discovery, 2019, 1(1), FDD11.
3. Armstrong, D. et al. Predictive Toxicology: Latest Scientific Developments and Their Application in Safety Assessment. In Comprehensive Medicinal Chemistry III. Elsevier, 2017, 94-115.

By Marie-Aude Guie, Vice President of Scientific Computing and Data Science

Artificial intelligence (AI) is now recognized as an indispensable component of the modern drug discovery tool kit, applied at all drug discovery stages from initial hit finding to lead optimization and beyond. It has been repeatedly shown that pairing high-quality data with advanced AI can dramatically enhance the hit finding process.^1,2,3 But as the benefits of AI in drug discovery continue to mount, we must remember that maximizing this industry’s potential means aiding key contributors in their efforts:

In short, if we want the AI community to continue supporting the drug discovery community, we must support the AI community too.

Think of it this way: AlphaFold can now predict 3-D protein structures with greater accuracy than ever.⁴ Why? Because training data were made publicly available. So, how do we further improve AI algorithms? We share data, empowering data scientists to explore new approaches to solving drug discovery challenges.

Unfortunately, most companies do not share their proprietary data, resulting in public datasets that are small and lacking in diversity. Additionally, machine learning (ML) requires both positive and negative data for effective models, but public (e.g., literature) datasets often only include positive results. This will not do.

To help correct the issue, X-Chem is taking steps to lead by example and be part of the solution: In collaboration with Structural Genomics Consortium (SGC), X-Chem will be publicly sharing its DNA-encoded library (DEL) screening data. AI practitioners will have open access to these data and be able to use them to test novel AI approaches. Further details are available in this press release.

Our hope is that X-Chem’s high-throughput, high-quality data will similarly drive advancements in AI, as AI has already enhanced various stages of drug discovery. With these new public datasets, we will also promote target validation via probe identification and explore previously unchartered avenues of drug discovery.

Our mission is to help our clients cure disease and share our innovations with the world. We’re already putting our technology in the hands of drug discovery experts, and now we aim to do the same for the AI community by putting more data in their hands. X-Chem’s DEL platform provides an unprecedented depth and breadth of clean and diverse data. And now these data will be available to the public.

References

1. McCloskey, K. et al. Machine learning on DNA-encoded libraries: a new paradigm for hit finding. Journal of Medicinal Chemistry, J. Med. Chem. 2020, 63, 16, 8857–8866.
2. Ahmad, S., et al. Discovery of a First-in-Class Small-Molecule Ligand for WDR91 Using DNA-Encoded Chemical Library Selection Followed by Machine Learning. J. Med. Chem. 2023, 66, 23, 16051–16061.
3. Li, A.S.M. et al. 2023. Discovery of Nanomolar DCAF1 Small Molecule Ligands. J. Med. Chem. 2023, 66, 7, 5041–5060.
4. Bordin, N. et al. 2023. Novel machine learning approaches revolutionize protein knowledge. Trends in Biochemical Sciences, Trends in Biomedical Sciences. 2023, 48(4), 45-359.

By Matt Clark, President & Chief Scientific Officer

Most readers will be familiar with the parable of the “drunkard’s search”, in which a drunk person searches for their lost car keys under the glow of a streetlight, not because it is close to where they lost the keys, but because “this is where the light is.” This story is commonly cited in the context of observational bias in the sciences, where one’s training and research tool kit leads them to frame problems in ways that may not be useful for the phenomena at hand. This concept is also captured in the proverb “if all you have is a hammer, everything looks like a nail.” In essence, our tools frame the way we study, explore and ultimately draw conclusions about the world.

I think of hit identification as the ultimate streetlight problem. We know that chemical space is incredibly vast, but our ability to sample it is limited. It is a testament to the diligence and creativity of the screening community that we can find actionable chemical matter from screening merely a few million compounds, considering the mind-numbing numbers of potentially useful compounds that exist in theory (i.e., ten to the power of 60, by one estimate). In my analogy, the area lit by the drunkard’s streetlight represents the size and composition of typical screening collections, and the unlit regions of the street symbolize the remaining untapped vastness of chemical space.

So, what’s the significance? Hit rates of modestly sized screening collections are sufficient to drive successful hit generation in most cases. But what additional value might lie in the “unlit” portions of chemical space? This question occupies many in the DEL community. Our libraries tend to be much larger than typical screening decks, so to continue the analogy, our lit area is much bigger. Does searching larger areas of chemical space yield benefits? Will we discover new ways to ligand proteins? Or will we merely find variations of what we’ve already seen — the same binding problems solved in slightly different ways?

A recent paper published by AZ and X-Chem explores the consequences of searching vaster areas of chemical space. In this Perspective, Collie et al., survey the literature for successful DEL screens where the binding interaction is characterized, typically by X-ray crystallography. Their findings reveal that, in numerous cases, the nature of the protein–ligand interactions is novel, compared to previous discoveries. New protein conformations, previously unknown binding sites and unprecedented interactions are frequently observed. Even for well-studied classes like kinases, the reported DEL hits often present novel aspects in their binding. The authors also demonstrate that these new interactions are not just academic curiosities; they seem to lead directly to highly selective ligands, even for kinases.

Unfortunately for the preparation of this manuscript, what may be the most powerful demonstration of this trend was published too late to be included. In a recent paper published in AACR Cancer Discovery researchers from Relay Therapeutics describe the discovery of a wild-type sparing, pan-mutant selective, allosteric inhibitor of PI3Kα using DEL screening. Notably, structural characterization demonstrated that the compound bound to a novel cryptic pocket, yet again illustrating DEL screening’s power to uncover unprecedented binding modes. This paper also tracked the hit through optimization and into clinical trials, where it demonstrated patient benefits due to its unique selectivity and binding properties.

The conclusion we draw from these observations — at the risk of overextending the analogy — is that there is indeed substantial untapped value in the dark areas of the street. It is there that we find compounds with behaviors distinct from those discovered in the past. In an era of tackling increasingly challenging new drug targets, a bigger streetlamp will prove to be a valuable tool.