Drug discovery paradigm could shift with cancer center involvement

Collaborations between data and computation-driven systems biology companies and forward-thinking oncology clinics are poised to change the face of cancer drug development and cancer itself.

Thomas A. Neyarapally
Faced with mounting productivity challenges, thepharmaceutical industry is in dire need of game-changing approaches todiscovery and development, and a new source of compounds to feed itsdiminishing pipeline. Despite the promise of drugs such as Gleevec andHerceptin, the current success rate in oncology drug development isparticularly unfavorable.1 The completion of the Human GenomeProject and subsequent advances in the molecular profiling of biologicalsystems (generating genomic, proteomic and other types of data) have drivensome successes in oncology, including discoveries of many important new genesand genetic mutations, and diagnostic tests such as Genomic Health's OncotypeDx test, but oncology drug development as a whole does not appear to haveachieved the major advances expected. Drug development productivity generallyhas not increased over the past several years, despite major ongoing R&Dinvestments in the latest technologies and increases in the quantity andquality of relevant data.2 

Pharma companies have historically worked with cancercenters with a focus on clinical trials, although cancer centers around thecountry continue to produce interesting new discoveries in oncology. In lightof the rapid advances in the cost, quantity and quality of measurementsavailable directly from patients, this is beginning to change, and a newparadigm is emerging. Today, these centers are uniquely positioned to become anabundant source of actionable knowledge for the drug development industry.Increasingly rich, high-quality data directly from patients is becoming morereadily available at reasonable cost, and thus higher volume, and is accessiblethrough the normal course of treatment through profiling of blood and tumortissue. Unfortunately, pharmaceutical and biotech companies have not yet foundeffective ways to fully access and leverage this rich genomics data collectionto accelerate their discovery and development efforts. This gap stems from aninability to scalably convert the explosion of "omics" and clinical data thatcan be gathered from oncology patients into actionable knowledge.
One solution is to find a technical bridge that connectsthe gap between the data and clinical and research expertise in cancer centersand pharmaceutical and biotech companies, enabling the conversion of the datagathered from cancer centers into actionable knowledge for drug discovery anddevelopment. Computational biology companies with the ability to take in dataat scale are the best positioned to do this, utilizing the emerging datamodalities of the new millennium—"omics" data (e.g., genetics, gene expression,proteomics, metabolomics, etc.)—to make novel drug and disease discoveries.Unfortunately, some critical missing pieces have forestalled the promisedresults of computational biology, including the lack of adequately powerfulanalytical tools capable of handling the increasingly rich and voluminous databeing generated, coupled with an inability to learn coherent models acrossseveral data modalities at once. Fortunately, by combining the best availablepowerful data-driven model learning and simulation algorithms with the latestavailable supercomputing horsepower, actionable knowledge can be extracted fromthis data.
Withthis new paradigm, there is another opportunity emerging. Cancer centeroncologists involved in the generation of the data could be in the bestposition to interpret the outputs of the models built from such data, with theassistance of the computational team. Further, while pharma companies are, bybusiness necessity, primarily focused on completing a trial and obtaining the resultson their drugs for regulatory approval, clinicians continue to provide care fortheir patients and can leverage the learnings from the computational analyseson an ongoing basis in administering that care. Clinics are also incentivizedto do all relevant research and explore all angles with respect to identifyingthe best treatments for patients, not just the angles that result in favorableinformation about a particular therapeutic.
Cancer clinics have key capabilities and resources that canbe brought to bear on the drug discovery and development problem. The requisitecomplementary capabilities and resources can be found in systems biologycompanies with approaches that are flexible enough and powerful enough to learndirectly and solely from previously unseen data, and to integrate "known"biology with novel discoveries identified in a data-driven way as needed. Suchgroups can work with several pharma and biotech partners without fear ofoverlapping results, as the models learned that reflect proprietary pharmainformation are learned directly from datasets of interest to the pharmacompany. Given the high dimensional nature of the datasets involved (e.g.,millions of SNPs, thousands of genes or other molecular entities and severalrelevant endpoints), the ideal companies for such partnerships would havetechnology platforms that are optimized to leverage supercomputing resources sothat results can be obtained in relevant time scales—hours or days, rather thanmonths or years.
Collaborationbetween truly data-driven systems biology groups and oncology clinics canresult in the ideal modern-day research tool—a simulation model that can allowresearchers to postulate a myriad of possible changes in a cancer and obtainanswers and confidence levels in those answers as to how the outcomes for thesystem (e.g., tumor size, survival, etc.) change as a result. Unlike a specificdiscovery (e.g., gene 1 appears to be associated with tumor size), such modelscan be utilized simultaneously by many groups to make many differentdiscoveries; these models make the "signal" inherent in a particular datasetreadily accessible to many researchers. For example, a model of glioblastomathat connections genetic variation in tumors to the biological processesunderlying the disease (e.g., gene expression networks) to outcomes such assurvival and recurrence, might be used to identify new targets for glioblastoma(by modulating different genes in the model and identifying which gene or genesmost strongly influence outcomes), and could also be used to stratify patientsfor a drug already in the clinic.
Thesemodels are similar to the gene maps and other research being produced by cancerresearchers,3, 4 except they are interactive and quantitative,enabling the researcher to ask many specific questions of interest, in realtime, and to obtain the answers quickly. In particular, the models that reflectgenetic variation will enable patient-specific recommendations via genotypes.
Collaborationsof the type cited above also have the benefit of rapid follow-up on both thecomputational and clinical sides. Discoveries made in the collaborations willhave already included clinicians who do not then need to be convinced of theapproach and the value of the results; they will be most readily disposed tousing drugs and diagnostics emerging from the approach at bedside. Further, tothe extent that the initial models indicate that more data of a certain typeare required in order to build truly predictive models, the relationship is alreadyin place for the clinical group to obtain more of the appropriate data.
In anincreasingly computationally driven world, the seamless interaction betweencomputation and the clinic is inexorable. These unique types of collaborationsare an opportunity to create the framework for and usher in a paradigm shift indrug research and development. This collaborative work will also create thefoundation for personalized medicine approaches to identify the righttreatments for the right patients, using the same types of modeling approaches.The data reside in the clinics, the computational platform and ability toleverage the power of supercomputers lay in the next-generation systems biologycompanies and the ability to turn novel targets and biomarkers into new, betterand more targeted oncology drugs lay in the pharma and biotech companies.
Forthese reasons, collaborations between data and computation-driven systemsbiology companies and forward-thinking oncology clinics are poised to changethe face of cancer drug development and cancer itself.
ThomasNeyarapally is responsible forbusiness development, corporate development and the expansion, protection andmonetization of Gene Network Sciences' intellectual property. He was appointedto his current position in 2008, having served as GNS' vice president ofcorporate strategy and intellectual property since 2006. Previously,Neyarapally served as an associate in the New York office of the law firmFrommer, Lawrence & Haug LLP, where he focused on transactional, productdevelopment and litigation matters in the pharmaceutical and biotechindustries. Neyarapally previously was an associate in the corporate departmentat Chadbourne & Parke LLP, and held the position of analyst at Arthur D.Little. Neyarapally holds a J.D. and an M.B.A. from Cornell University. Whileattending Cornell's Johnson Graduate School of Management, he served as apartner with BR Ventures, the only student-run venture capital firm in theUnited States. Neyarapally graduated from the University of Connecticut with aB.S. in chemical engineering.


1.Ismail Kola and John Landis, "Can the Pharmaceutical Industry Reduce AttritionRates?" Nature Reviews: Drug Discovery3(8) (August 2004): 711-715.

2.Ismail Kola, "The State of Innovation in Drug Development," ClinicalPharmacology and Therapeutics 83 (2008):227-230.

3. LiDing, et al., "Somatic Mutations AffectKey Pathways in Lung Adenocarcinoma," Nature 455 (2008): 1069-1075.

4.Markus Bredel, et al., "A Network Modelof a Cooperative Genetic Landscape in Brain Tumors," JAMA 302(3) (2009): 261-275.

Thomas A. Neyarapally

Subscribe to Newsletter
Subscribe to our eNewsletters

Stay connected with all of the latest from Drug Discovery News.

Front Cover

Latest Issue  

• Volume 18 • Issue 12 • December 2022/January 2023

December 2022/January 2023

December 2022/January 2023 issue