Smoothing the way for personalised medicines

The right tools can help scientists identify and develop targeted therapies more quickly and cost-effectively. Dr Sharon Benzeno, Senior Director, Business Development Bioinformatics at Reed Elsevier, considers where the industry goes from here

Extracting actionable information from vast volumes of large data sets is a major bottleneck hindering the advance of precision medicines

Before personalised medicine, most patients with a given disease received the same treatment, even though many didn’t respond. Today, greater knowledge of genetics and genetic susceptibility has enabled the targeting of treatment plans to each patient’s needs, particularly in cancer. Targeted therapy is directed at a cancer’s specific genes, proteins or micro-environment that are known to contribute to its growth and survival. Examples of classes of targeted drugs include hormone therapies, gene-expression modulators, signal transduction inhibitors, angiogenesis inhibitors, immunotherapies and apoptosis inducers.

We’ve made progress in developing more effective, targeted treatments and determining subpopulations of responders, but more work is required

At the same time, pharmacogenomics has the potential to reveal whether a targeted therapy is likely to work in a particular patient by looking at how the person’s genes affect the way the body processes and responds to a specific drug; this, in turn, influences drug safety, efficacy and dosing for that individual. Breast, kidney, lung and colorectal cancer are among the cancers for which targeted treatments are available for the right patients. Targeted therapy also is becoming available for other conditions, such as rheumatoid arthritis and certain forms of diabetes – and many more targeted options are needed.

However, targeted therapies are not without some downsides, as they may cause side-effects or stop working altogether (e.g. a tumour becomes resistant to therapy); genetic testing of both the patient and, in the case of cancer, the tumour, can be complex and costly. The treatment itself can, on average, reach upwards of US$1,000 a month.1 In short, we’ve made progress in developing more effective, targeted treatments and determining subpopulations of responders, but more work is required.

Making the most of data

Next-generation sequencing and whole genome profiling platforms – the technologies that underpin much of personalised medicine – are becoming increasingly affordable, leading to an enormous influx of potentially valuable data to help make better drug development and clinical decisions. In parallel, patient-centric data from various sources – e.g. electronic health records, claims, pharmacy and health insurance profiles, medical administration records and mobile diagnostic and monitoring devices – is also becoming available to help inform treatment decisions. Add data from clinical trials and mechanistic studies to the mix, and the major bottleneck facing personalised medicine emerges: extracting actionable information from vast volumes of large data sets.

Many data-mining tools are able to process only certain types or aspects of data, potentially missing critical input, and thereby skewing interpretation

Today, many data-mining tools are able to process only certain types or aspects of data, potentially missing critical input, and thereby skewing interpretation. This may lead to incomplete coverage of information, low reproducibility of clinically relevant assessments, and even disagreement regarding the significance of the outputs.2 There is a need to better mine and interpret large volumes of data from diverse sources to draw appropriate clinical conclusions and to bring effective targeted therapies to market. Indeed, the most current, accurate and comprehensive tools are needed to manage these specific data-related challenges.

1. Structure and quality: More than 85% of medical data is clinically relevant, but unstructured.3 Normalising that data to mine it effectively is a complicated process that requires sophisticated tools. Also, differences among sequencing platforms, and the presence of sequencing artifacts4 and other anomalies can compromise data quality and analysis.

2. Actionability: Identification of altered genes and proteins associated with a specific disease can lead to early detection and help guide treatment options. Still, more data is needed to make better informed decisions. In addition, most patients do not participate in clinical trials (only 2% to 3% of cancer patients who are eligible for clinical trials participate5,6). The ability to extract and analyse data from ‘real-world’ patients could lead to new targeted treatments and refine the subpopulations of responders versus non-responders.

3. Standardisation: Uniform analytical protocols are vital if R&D is to move more efficiently and reliably from data generation to data analysis. Data standardisation eventually will allow organisations to integrate information from multiple and diverse sources, thereby providing more accurate, substantive outputs to R&D queries. Standardised data also will enable effective R&D collaboration by facilitating seamless information exchange and utilisation among internal and external stakeholders.

4. Curated content: Accurate interpretation of data requires access to high quality curated content that enables both scientists and healthcare providers to quickly and reliably assess the most up-to-date information about genetic variants and associated phenotypes from sequencing and biomarker data. This information, which can inform both drug discovery/development and clinical decision-making, is best derived using the most comprehensive data extraction, aggregation and integration technologies.

Getting to targets faster

New treatments need to be aimed at the right targets. Target selection requires mining diverse sources of evidence, including experimental, clinical and published peer-reviewed data. The most reliable analytical solutions process multiple types of data sets (e.g. gene expression, proteomics, protein-protein interactions, cell processes, disease mechanisms, functional drug classes, etc.) that contain key concepts, such as ‘gene’, ‘protein’ or ‘drug’. Automated entity recognition and pattern matching software is utilised to identify meaningful relationships between targets and molecules/biologics (e.g. ‘binds to’, ‘activates’, ‘inhibits’).

The same technologies and processes used to identify novel targets can be highly effective in identifying potential new indications for already approved drugs

Integrating the extracted data into a genomic, proteomic and biomarker knowledge base not only enables the identification of new target discoveries; it also enables R&D to investigate potential new indications in areas of unmet medical need for existing drugs or drug candidates within a company’s pipeline.

The latter is important as pharmaceutical companies increasingly turn to drug repositioning and repurposing to recoup return on investment and to rapidly bring new treatments to market—and ultimately to patients—with minimal risk (i.e. drugs that have been approved or passed through Phase I studies have already undergone safety/toxicity testing).

The same technologies and processes used to identify novel targets can be highly effective in identifying potential new indications for already approved drugs

The same technologies and processes used to identify novel targets can be highly effective in identifying potential new indications for already approved drugs. Since many approved targeted therapies are in fact ‘dirty’ (not purely selective for a single target), the ability to reposition them in another disease or indication (drug-centric repositioning) is valuable. Similarly, a ‘failed’ drug may emerge as a potential therapy for a disease for which there is no effective drug on the market (disease-centric repositioning). Various methods7 are being used for drug repositioning, all of which generate data that need to be analysed, aggregated and integrated in the appropriate contexts.

Targeted mechanism-based: Create mechanistic therapeutic models using treatment ‘omics’ data analyses.

Signature-based: Explore functional connections between biological events using an enhanced gene-signature design, assisted by pathway and network analyses.

Pathway- or network-based: Reconstruct disease-specific pathways using disease ‘omics’ data analyses.

Knowledge-based: Provide comprehensive bioinformatics/ cheminformatics analyses of available information from drug-target networks, chemical structures of drugs and targets, clinical trials and other input.

Drug repurposing/repositioning programmes applying some of these approaches are providing new treatment options for patients while cutting an estimated two to three years and about $10m off the approval process.8 A few successful examples of drug repurposing include:

Thalidomide A derivative of glutamic acid, thalidomide was originally developed to treat nausea during pregnancy and was withdrawn from the market because of its teratogenic effects. One of the proposed mechanisms for those effects was the drug’s anti-angiogenic activities. Those activities were further investigated and led to the successful reintroduction of the drug to treat cancer. Thalidomide is now approved for the treatment of multiple myeloma, and recent studies suggest efficacy against additional malignancies, including myelodysplastic syndrome and acute myeloid leukemia.9

Drug repurposing/repositioning programmes are providing new treatment options for patients while cutting an estimated two to three years and about $10m off the approval process

Mitoxantrone This toposomerase II inhibitor is used in the treatment of metastatic breast cancer and leukemia. Recently, researchers screened 89 drugs already approved by the US FDA in an effort to identify additional therapies to treat gastrointestinal stromal tumors (GIST).10 Mitoxantrone emerged as one of two promising compounds (the other is the gene transcription inhibitor mithramycin A, an antineoplastic antibiotic currently under investigation for the treatment of Ewing sarcoma) to treat GIST.

Duloxetine Originally approved to treat major depression, this serotonin and norepinephrine reuptake inhibitor was subsequently also approved to treat musculoskeletal pain and fibromyalgia.11

Suramin12 First used almost a century ago to treat African sleeping sickness, suramin – which has an unknown mechanism of action13 – is now being studied as a potential treatment for, among others, kidney disease, liver disease and autism.

These and other successes have fuelled numerous collaborative approaches to drug repositioning. Among the most ambitious is a project of the US National Center for Advancing Translational Sciences (NCATS). Its ‘New Therapeutic Uses’14 programme brings pharma companies together with biomedical researchers who test contributed compounds for additional indications.

NCATS provides a table of ‘industry-provided agents’ to facilitate exploration of potential new therapeutic uses.15 The available agents have undergone pre-clinical, safety and dosing testing and are ready for additional testing in humans.

The FDA also has compiled ‘The Rare Disease Repurposing Database’,16 which consists of products that have received orphan status designation and are already market-approved for the treatment of other diseases.

A recent report17 by Dr Minna Allarakhia, director of the nonprofit BioEndeavor, provides a roundup of open-source initiatives for repositioning drugs for neglected as well as common diseases, and includes a number of collaborative efforts around the globe.

All these efforts reinforce the need for and impact of personalised medicine: to provide more treatment options for subgroups of the larger population – those with rare diseases; those with more common illnesses who don’t respond to conventional treatment; and those with genetic variations whose illnesses are more likely to respond to specific targeted therapies.

Going forward, a better understanding of mechanisms of action will further enable the development of new drugs and the repositioning of others to expand the applications of personalised medicine and refine treatment options by disease and patient type.


1. Langreth R., Bloomberg, May 7, 2014.

2. Dewey et al., JAMA Vol 311, No. 10

3. Demystifying Big Data: A Practical Guide to Transforming the Business of Government. Prepared by TechAmerica Foundation

4. Zhou et al., PLOS One, April 02, 2013. DOI: 10.1371/journal.pone.0060234

5. Cancer Research Institute: About Clinical Trials

6. Why are Only 3% of US Cancer Patients in Clinical Trials? Medscape Oncology. January 13, 2013.

7. Jin G and Wong STC, Drug Discovery Today. January 2014.

8. McKew JC. Drug Repurposing at NCATS

9. Cancer drug discovery by repurposing: teaching new tricks to old dogs

10. UBoichuk et al., Cancer Research. Published OnlineFirst January 2, 2014;
doi: 10.1158/0008-5472.CAN-13-1955

11. Li YY and Jones SJM., Genome Med. 2012; 4(3): 27. Published online Mar 30, 2012.
doi: 10.1186/gm326

12. Bloom B. Repurposing is Hidden in Plain Sight!

13. DrugBank: Suramin

14. NCATS: Discovering New Therapeutic Uses for Existing Molecules

15. NCATS: Industry-Provided Agents

16. FDA: Rare Disease Repurposing Database

17. Allarakhia M., Des Devel Ther. 2013; 7: 753–766. Published online Aug 8, 2013.