As a multidisciplinary activity, drug discovery relies on a range of life science analytical and processing platforms. Separation, quantification, and characterization use liquid and gas chromatography, capillary and gel electrophoresis, surface plasmon resonance, and even viscosity and metals determinations.
Of all the detection modes available today, mass spectrometry (MS) is arguably the most sensitive and versatile. At a basic level, MS tells medicinal chemists whether they’ve made the molecule they want and which side products, degradation molecules, or impurities are present in the drug substance or drug product. MS has long been used to investigate a drug’s metabolism and pharmacokinetics in both vitro and in vivo pharmacology.
With the advent of the ’omics disciplines and their nearly immediate impact on pharmaceutical R&D, MS now finds itself at the forefront in the study of biomarkers, primary and secondary metabolites, and other large and small molecules relevant to understanding how medicines work.
All the while, MS developers have kept pace with pharmaceutical industry needs. “For a long time, triple quadrupole MS was the standard of analysis, but over the last five years, higher-end instruments have become commonplace, particularly TOF (time-of-flight) and quadrupole TOF spectrometers that are capable of both qualitative and quantitative work,” comments Diane Diehl, PhD, director of pharmaceutical market development at Waters (Milford, MA). “These instruments quantify what happens to drugs after they’re taken and can even characterize unknowns. Drug discovery scientists can now get more and richer information from one injection instead of conducting multiple studies.”
Analytics in discovery/development settings must deal with high background components in the sample matrix and very high concentration dynamic ranges. No modality achieves this better than MS, particularly for unknown species whose detection demands sensitivity. Waters has also been investigating MS techniques including ion mobility and collisional cross-section (CCS). Ion mobility mass spectrometry (IMMS) technology rapidly separates molecular ions by not only their size and mass-to-charge ratio but by their shape as well and calculates a CCS value, a physicochemical property unique to each molecular ion. Getting a CCS value for each ion reveals insights into the chemical structure of sample molecules, thus providing a higher degree of specificity than gained from relying on a mass-to-charge ratio alone.
INFORMATICS TO THE RESCUE
Waters is upgrading informatics as well through improvements in its own MS data software, UNIFI Scientific Information System, through work with third-party informatics companies, and through acquisitions of software companies.
Informatics, Diehl says, has become as significant in drug discovery as the analytical method itself. “Especially for the analysis of biomarkers and in metabolomics, for example, comparing control and affected patients, treated and untreated, and comparing their metabolomic fingerprints in a meaningful way.” This could take the form of a peptide map or a pure metabolomics study, for example. Small-molecule analysis is a specialty of Waters’ Progenesis QI software, which processes small-molecule metabolic profiling data. Its SONAR data acquisition and Symphony software products manage discovery proteomics, lipidomics, and metabolomics.
High sample numbers and high-resolution readings add linearly to the data pileup, but as the science becomes more complex, drug discovery experiments become studies within studies. Known and unknown peaks are matched to public or proprietary mass databases. Expert data processing by hand (or with the aid of a computer), moreover, is becoming obsolete, as discovery companies tend to hire scientists with broad experimental design, execution, and interpretation skills. “Interpreting an experiment manually can take weeks and is often the source of error. Software eliminates those mistakes,” Diehl says.
While error reduction is less of a regulatory issue in discovery than in manufacturing, formulation, and packaging, a focus on data integrity during early-stage work cannot hurt, and it may in fact help anticipate quality issues later in a product’s life cycle.
Immuno-oncology, an emerging therapeutic modality, is an area where discovery can use all the informatics assistance it can get.
Immuno-oncology treatments harness the body’s own healing mechanisms to attack and eliminate cancer. Passive immunotherapies use conventional cancer-fighting drugs (e.g., antibodies, cytokines) to enhance anti-tumor responses. Active immunotherapy involves harvesting a patient’s own immune cells, reprogramming them to enhance anticancer activity, and reinfusing them into the patient. Developing these treatments involves exquisitely balancing beneficial and adverse immune mechanisms, which is why they work spectacularly in 10 to 20 percent of patients and fail in the rest.
In late 2016, Pfizer (New York, NY), a leader in immuno- oncology, teamed up with IBM Watson Health (Armonk, NY) to identify and promote novel immuno- oncology treatments. Together, they will leverage IBM’s cognitive computing tool, Watson Drug Discovery (WDD), with Pfizer’s deep biological expertise. The collaboration is a first for both companies.
“Immuno-oncology is not a one-size-fits-all approach,” says Jadwiga Bienkowska, PhD, senior director of computational biology, oncology translational research at Pfizer. “The future of immuno-oncology lies in novel, biologically rational combinations tailored to unique tumor characteristics that enable treating more patients and ultimately transform cancer treatment.”
Bienkowska’s “combinations” refers to the use of multiple agents, including one or more standard immune-stimulating agents, delivered through a personalized or semipersonalized strategy. Pfizer is developing approximately 10 immunooncology treatments targeting a variety of pathways.
WDD is a new cognitive computing tool tailored for drug discovery. It rapidly analyzes and tests hypotheses to generate evidence-based insights and dynamic visual representations that researchers can interact with in real time. WDD also supports drug safety investigations.
LEVERAGING DISEASE MODELS, INDICATIONS
Drug discovery involves testing to validate targets and screen compounds. At some level, test throughput becomes an issue. Whether experiments involve biochemical tests, cell-based assays, animal studies, or human trials, the object is to glean as much information as possible as economically and scientifically rigorously as conditions allow.
Advances in molecular biology have given rise to drug discovery methodologies that rely on the interaction of putative drugs with disease-related targets or chemical surrogates to investigate efficacy and toxicology. Target- based drug discovery, or “reverse pharmacology,” has been the predominant drug discovery paradigm, but as we will see, it is not entirely justified. A survey reported in a 2011 paper by Roche scientists David Swinney and Jason Anthony found that 62 percent of approved new first-in-class drugs arose from phenotypic screening despite the prevalence of target-based approaches. The authors concluded that “... a target-centric approach for first-in-class drugs, without consideration of an optimal MMOA [molecular mechanism of action], may contribute to the current high attrition rates and low productivity in pharmaceutical research and development.”
The theraTRACE® phenotypic discovery platform, developed by Melior Discovery (Exton, PA), screens druglike compounds using in vivo assays in more than 40 validated animal models, spanning multiple therapeutic areas: inflammation, immunology, diabetes and metabolic syndrome, dermatology, cardiovascular, gastrointestinal, psychiatric, and neurological and neurodegenerative disorders.
The animal models are identical to those used by pharmaceutical companies but are assembled in a proprietary “multiplexed” format that rapidly and cost-effectively evaluates compounds for efficacy in multiple therapeutic areas. A complete pharmacological profile emerges in about 10 weeks.
“theraTRACE makes practical what is normally impractical in a drug screening program,” says Melior Discovery president and CEO Andrew Reaume, PhD. “Each rodent test animal cohort allows us to perform three or four times as many assays as normal.”
For example, a metabolic disease animal cohort model with an innate predisposition to particular pathologies may be put on an experimental diet for eight weeks and administered, say, a previously failed Alzheimer’s drug. In addition to evaluation for the effects in weight gain, it may be possible to administer a hot-plate test to evaluate that same animal cohort, using the same drug, for analgesia.
“Normally, one would worry about assay-assay interaction,” Reaume notes. “theraTrace institutes appropriate sequences of tests that can be used together without compromising the quality of either test.”
Melior’s brand of phenotypic testing complements target- based discovery, with the potential to shorten discovery/ development timelines that currently run as long as 12 years. It also provides a mechanism for testing compounds that have already been proven safe but failed to reach one or more endpoints for clinical efficacy. Thousands of such compounds exist at pharmaceutical companies. Many are highly “druggable,” meaning their solubility and permeability are adequate for conventional formulation. These compounds are, in effect, candidates without indications. “These are highly privileged assets with good tolerability profiles in humans. Drug developers have already invested tens of millions of dollars in them,” Reaume says.
He cites sildenafil (Viagra) as an example of a highly successful drug that was “discovered” due to serendipity—a diversion from the track of the original hypothesis for efficacy.
Originally developed to treat angina, sildenafil dilates cardiac blood vessels by inhibiting phosphodiesterase-5 and acting on the messenger molecule cyclic GMP. Pfizer, the developer, did not find significant angina relief in human testing, but observant clinicians noted that some patients developed penile erections. Meanwhile, academic researchers gathered evidence that nitric oxide, whose formation is mediated by cyclic GMP, was necessary for penile erections. Putting two and two together, Pfizer switched development gears, and a new drug for a new indication was born. Viagra has become the poster child for sound molecular biology in search of a useful phenotype.
Reaume estimates that as many as 90 percent of compounds that the FDA ultimately approves succeed based on some element of serendipity that took researchers off the development track they started on. “Many of our pharmaceutical industry partners have realized that they don’t have to wait until a compound fails in clinic to uncover otherwise unpredicted biology. We have shown that they can succeed by exploiting and systematizing serendipity.” Many of Melior’s current projects involve novel targets or molecule classes, but with the understanding that more biology may exist than was anticipated. “It’s best to uncover that earlier than later,” Reaume adds.
“It’s not that a small-molecule drug is being promiscuous and simply hitting another target. The original target is what leads to the discovery. We just never appreciated or anticipated all the biology that was linked to that target.”
Phenotypic testing can be extended to in vitro assays that similarly put observations before hypotheses. We see this today primarily with cell-based assays, particularly with organoids that mimic physiologic tissues more closely than individual cells. According to Reaume, this activity has “exploded” during the past decade. “It’s much bigger and more diverse than animal models. It’s now possible to probe for various cellular outcomes and deconvolute the biology afterwards.”
Yet Melior is sticking with its multiplexed animal models. The company has been approached by computational biology companies that seek to validate in rodents the complex connections they have established between targets, proteins, and drugs in vitro. Computational drug discovery makes sense of myriad molecular interactions, but as a predictor of efficacy and safety, it is only as good as its data inputs.