Modern drug discovery begins with a target—a molecule that participates materially in one or more pathways involved in disease maintenance or progression. Drug developers reasonably assume that interfering with target’s function pharmacologically will interrupt disease-associated pathways and improve the patient’s health.
More targets are generally available for biologic drug development but even then their number is limited. “The goals of target validation are similar for small molecules and biologics,” comments Joanna Lisztwan, EVP Head of Global In Vitro Pharmacology at Evotec. “In both instances you want to ensure the desired therapeutic outcomes can be achieved through the targeted mode of action.“
Biologic drugs might open up more possibilities, e.g. for targets that are inaccessible to or are unaffected by small molecule drugs—and the reverse is true as well. “However, as long as the validation process follows a clear line of scientific questioning, and applies the appropriate technologies correctly, drug modality should be agnostic to target validation,“ Dr Lisztwan says. “Generally speaking, understanding the desired drug mechanism of action is more important for target validation than the therapeutic modality. The question whether the targeted mechanism requires an agonist or an inhibitor fundamentally shapes the experimental design of the validation process.”
Is it druggable?
Dozens of potential targets may exist for a disease but their suitability to pharmacologic development depends on their “druggability.”
In their review of target validation, Gashaw et al defined a druggable target as a molecule whose activity is modified pharmacologically, is quantifiable through assays—preferably high-throughput screens, expressed preferentially in the diseased tissue to limit side effects, and have associated biomarkers suitable for monitoring treatment or disease progression. And above all, target and associated assays should be free of intellectual property restrictions.
The thoroughness of target validation depends on the quality of data supporting it. “Drug developers should base their go/no-go decisions on data generated through exceptional scientific and operational excellence,“ Dr Lisztwan says. “Developers must decide early on whether the disease biology or phenotype is reflected best through an in vitro or an in vivo model. Once they settle on the model, they face many choices for method, including tool compound, knock-out, knock-down, knock-in, overexpression, etc.” The most robust method is selected on the basis of the ability to reliably quantify changes in disease-related biological processes. “The decision to move ahead with a drug discovery project is possible only if the applied technology or assay method delivers reliable results in tracking target expression and target manipulation.”
Enter artificial intelligence
Much has been written regarding drug discovery workflows and workloads. Automation mitigates issues of hands-on vs. walk-away time, consistency, and human error. Software handles voluminous data more efficiently than a post-doc with a spreadsheet. The final frontier in these ever more-complex workflows is the transformation of data into intelligence. Companies are now turning to artificial intelligence (AI) and machine learning (ML) to uncover hitherto inaccessible patterns in these data.
“AI provides several improvements over legacy approaches like ‘one-factor-at-a-time’. Through techniques such as deep learning, AI is able to transform diverse, high-volume, multidimensional datasets for a thorough interpretation of the total set of the data” says Ben Miles from Strateos. Among AI’s benefits are the identification of novel target and biomarkers, discovery of previously unknown disease-target associations, phenotypes, and disease mechanisms, and overall better design of small molecules and biologics.
Adopting AI or ML in target validation is not for the faint-hearted. As Miles explains, the pace of scientific discovery has accelerated far beyond the comfort zones of traditional drug discovery teams. “Keeping up with rapidly changing technologies and accumulated knowledge has become paramount. Laboratories need to expand outwards and incorporate scientific workflows, software, statistical programming languages, and best-in-class data tools. Partnering is one way to acquire these capabilities and compress the timelines from hypothesis to results.“ Moreover these efforts must capture all data and adhere to FAIR—Findable, Accessible, Interoperable and Reproducible—data principles to ensure generation of high-quality, consistent, well-structured data suited for AI or ML.
Through its remote-controlled laboratory platform, Strateos offers web-based, on-demand control of drug development (including target validation) services for both small molecule and biologic drug discovery. These include phenotypic and target-based screening incorporating cell-based, biochemical, and biological assays to support high-throughput screening, lead optimization, and structure-activity studies.
Risk vs. benefit
For the last 15 years the U.S. FDA has recommended a risk-based approach to pharmaceutical development and manufacturing to address potential issues in drug safety, efficacy, and manufacturing.
“Safety and efficacy are woven into streamlined methodologies for target validation,” explains Vad Lazari, Associate Director for Integrated Biology at Charles River. “For safety and efficacy you want the process to work backwards from patients to basic assays, which demonstrates target relevance. Signaling pathways are often up- or down-regulated in disease compared with healthy populations.” Since targets represent independent safety risks, their expression in multiple tissues suggests the potential for toxicity or side effects. “One should also consider signaling pathways downstream of the target.”
Patient populations also need to be considered from a risk perspective, as different demographics and treatment regimens present varying safety risks.
Validation eventually relies on in vitro assays on patient-derived materials conducted in the most humanized models possible, Lazari tells Biocompare. “These strategies create opportunities to identify suitable markers that change as the disease or treatment progresses, or which are suitable for monitoring safety risks. During target validation we can examine the effects of partial and complete target inhibition, as well as potentiation, to determine their effects on safety and efficacy.”
Assay developers need to consider a wide variety of factors when designing and validating target validation assays, including a need to balance throughput and physiological relevance, adaptability to laboratory automation, and, says Lazari, “a relentless focus on reproducibility.”
Most high-throughput screens test molecules at a single concentration, so the compounds have only one chance to show activity. Assays that lack reproducibility will lose potential hits, while a high false-positive rate can generate an unmanageable number of bogus ones. Developers manage assay inconsistencies by applying Zhang’s Z-factor, which provides a numerical value indicative of assay reliability.
Nitty gritty
Most small molecule libraries are stored in DMSO, a solvent that balances solubility and stability but which can also affect assay performance. DMSO may affect enzyme catalysis or poison cell-based assays, particularly those involving primary cells. “Nanoliter dispensing and production of assay-ready plates from high concentration compound stocks have mitigated solvent issues to some degree but it is still relevant,” Lazari says.
Assay developers therefore must establish their assays’ tolerance to DMSO before initiating a compound screen. “This can often be performed alongside assay development, where a known inhibitor is dissolved in DMSO and solvent-matched controls used to generate the uninhibited signal. A titration of DMSO into a developed assay is also recommended to establish an absolute maximal tolerance in case higher concentrations of compounds need to be tested,” Lazari explains.
Lazari also cautions that target validation efforts include consideration of timeframes during which an assay is expected to work. “Assay performance must be consistent or stable over time, over days, weeks, or months and across different operators, groups, and locations.” Developers achieve this by using a reference assay consisting of the test system and a molecule, known to work within those assay conditions, each time the assay is performed. “The variability of the data obtained with the reference compound, along with the underlying raw data values, is an indicator of assay stability. The more stable your assay the greater confidence the project team can have in the resultant data.”