Someday, it should be possible for doctors to send individual cancer patients in for a genomic analysis and, based on the results, prescribe the drug they know will be the most effective. While the promise of this kind of personalized medicine is still distant, researchers like Fátima Al-Shahrour, head of the Translational Bioinformatics Unit in the clinical research program at the Spanish National Cancer Research Center (CNIO) in Madrid, are working on it now, interpreting the genomes of individual cancer patients and searching for clues to how they will respond to various treatments. The field, which is known as cancer pharmacogenomics, is still in its infancy, but Al-Shahrour believes that "in the future [it] could benefit many people."

Trained as an organic chemist, molecular biologist, and bioinformaticist, Al-Shahrour sees many exciting opportunities for early-career scientists who are willing to work at the crossroads where biomedical research, bioinformatics and computational biology, and clinical research meet. "Medicine is going into that direction, so every hospital, every clinician, every laboratory in the future is going to need people who can interpret those results," she says.

From cancer mutations to personalized treatments

The CNIO research that Al-Shahrour is involved in starts when her clinical colleagues recruit cancer patients for whom conventional treatments have been exhausted. A small group of cancer patients are currently involved in CNIO's search for alternative treatment options, with a variety of cancers including melanoma, glioblastoma, and pancreatic cancer. Once they're recruited, experimental biologists perform a genomic profile of each patient, sequencing the exomes—the coding portion of the genomes—of individual tumors.

As the lead bioinformaticist on the team, Al-Shahrour’s role is to analyze the genomic data in search of mutations. She attempts to match the mutations with existing literature and databases to predict how a patient is likely to respond to nonconventional drugs. This work can, in turn, inform treatment decisions in the clinic.

Her work also feeds into an experimental approach to treatment, helping biologists decide which drugs to test in animal models of the patients’ tumors—xenograft mice used as proxies, or avatars, as the team calls them—to test how effective particular drugs could be in particular patients. The idea is that the treatments that the mouse responds to best can then be administered to the patient.

Beyond the treatment of current patients, Al-Shahrour is helping to develop a database of novel mutations and their associations to drug responses, together with computational methodologies, that could help predict drug responses in future patients. Now, "we find key mutations in genes that are expected to be mutated, but then we find many other mutations," which need to be tested in avatars to determine whether they might be clinically relevant. Eventually, she hopes to reach the point where, "if we find these mutations or any similar ones [in new patients] … we can give them a treatment that has already been given to another patient with a similar genomic profile."

The study is still in its pilot phase. Avatars have been successfully created for about half the patients, and some patients have been treated with nonconventional drugs that seemed promising in their avatars. "As a proof of concept, it has worked for a few patients," Al-Shahrour says—but more successes are needed to put the approach on firmer footing, she adds.

The stakes are high. For cancer patients who do not respond or relapse after conventional therapy, "there is no other treatment except to include them in a clinical trial or [offer them] this possibility that we are putting in place."

Special Issue: Cancer Genomics

A collection of articles from Science, Science Signaling, and Science Careers examines how a whole-genome approach is shaping our understanding of cancer.

At the heart of translation

In this multidisciplinary project, one role that Al-Shahrour plays is to ensure that genomic information reaches clinicians in a useful form. You can’t just give a medical doctor making treatment decisions a list of 3000 mutations found in those patients' tumors, she says. "Rather, you have to give him … five genes that are potentially important based on their clinical relevance."

Al-Shahrour also uses the interpretation of genomic data to bridge bioinformatics and biomedical research. Once a novel mutated gene has been identified in avatars, it's necessary to study its biological function to confirm whether "this is the gene [causing] this predisposition to be sensitive to this treatment," Al-Shahrour says. She contributes to the discussions and decisions about which experiments to prioritize.

Another aspect of Al-Shahrour’s job is to provide bioinformatics and computational support to clinicians and experimental biologists. She works at the interface of the computational biology and clinical research programs at CNIO, which puts her "in a unique, critical and challenging position," writes Alfonso Valencia, the director of the institute’s Structural Biology and Biocomputing Programme, in an e-mail to Science Careers. "Her group is responsible [for] organizing, digesting and analyzing the vast amount of data produced by the clinicians, including genomics and medical information, as well as the results of the analysis of xenografts. … A task for which she has the support of my group but also the implicit task of pushing us to streamline and optimize our tools and methods to fit the needs of the analysis."

But the biggest challenge for Al-Shahrour is the now very limited knowledge of the functionality of the genome. Finding mutations that you know neither the cause nor the effects of is frustrating. "[Y]ou know that only in the future will we have the methodologies to infer these functionalities and to be able to assign and interpret [the genome] at the clinical level," Al-Shahrour says.

A new breed of scientist

According to Valencia, the job that Al-Shahrour does requires a very wide range of knowledge and skills; he emphasizes her "biological background, capacity to develop bioinformatics methods, deep understanding of genomics, good communication skills and proved record in team management." Also important, he adds, is her clear understanding of the limitations of the experimental and computational techniques.

But what Al-Shahrour herself sees as her most important asset is her broad view of the field, encompassing the development of computer tools, databases, and computational methodologies and their use to study genes, cell lines, and patients. She developed this broad view via a series of career steps, first obtaining a B.Sc. degree in organic chemistry. She began her Ph.D. studies in molecular biology at CNIO in 2002, just as the use of DNA microarrays was becoming mainstream and methods were being developed for the functional analysis of genome-scale experiments. Under the supervision of Joaquín Dopazo, she worked on computational methodologies for microarray gene expression analysis, integrating databases and applying statistical tools to inferring which genes were most functionally relevant.

After her Ph.D., Al-Shahrour went on to work at the Broad Institute in Cambridge, Massachusetts, in the computational biology and bioinformatics lab of Chief Informatics Officer Jill Mesirov, where she worked closely with cancer computational biologists, one of whom was Pablo Tamayo. There, she "pioneered the use of molecular signatures to characterize the cellular state of cancer cells. This included projecting a variety of datasets in the space of genes induced by a variety of oncogenes," Tamayo writes in an e-mail to Science Careers. In October 2008, Al-Shahrour joined the lab of clinician-scientist Benjamin Ebert at Brigham and Women’s Hospital in Boston as a staff computational biologist, studying cancer biology and treatment using hematopoiesis as a model system. During her time in Cambridge and Boston, Al-Shahrour says, she learned how to work in large multidisciplinary groups, and clinical exposure taught her that, for bioinformaticists, the job isn't just to analyze the data but also to design the studies and interpret the data.

Early-career scientists who wish to follow in her footsteps must be ready to embrace the training challenges. Tamayo writes: "My advice to them is to study mathematics (not only old statistics but also advanced probability), information theory, machine learning, programming, numerical methods, chemistry, physics, cellular biology and biochemistry. It is important not only to be able to talk to multiple domain experts, and develop a solid hard-core analytical mind frame to cast problems, but also to have access to a rich set of paradigms about how to deal with complexity." Cancer pharmacogenomics is "a particularly demanding field that requires a lot of flexibility and adaptability in terms of what problems one solves over time and in requiring to learn from many fields of expertise," he adds.  

The challenges in the field are considerable, but Tamayo, Valencia, and Al-Shahrour all see great promise in cancer pharmacogenomics as an approach to treatment and as a career. "For the first time we are analyzing real data, that is, data from patients," Al-Shahrour says. She has moved from searching for mutations largely for the sake of knowledge to interpreting the genome to more directly help patients, which is, to her, the most exciting part of her research.

Elisabeth Pain is contributing editor for Europe.

10.1126/science.caredit.a1300050