Will sequencing prove the Beginning of the End, or the End of the Beginning for biopharma?
September 28, 2010 | Even for a jaundiced veteran of the biopharmaceutical industry, it’s hard not to gush over the astounding output of today’s DNA sequencing machines. Twenty-five years ago, I was lucky to read 1,000 nucleotides off an X-ray film during my thesis work. A next-generation sequencing (NGS) instrument can deliver nearly a billion times that number of bases. The computer sector, Moore’s Law notwithstanding, can’t match that gargantuan gain. But what is the potential value of this new technology to biopharma?
The pharmaceutical industry is notorious for its poor record of clinical prediction. When we spend tens of millions of dollars on a drug candidate with the expectation that we’ve discovered a safe and efficacious drug, fully nine times out ten, we’re wrong. Can massive amounts of DNA sequence data help?
Better biomarkers to guide therapy and new drug targets are areas of substantial need and opportunity. We should assume that within ten years a large fraction of patients will have their genome sequence available to assist doctors in the diagnosis and prevention of illness. Presumably, cancer patient reports will contain tumor DNA mutation analysis. Because the time frame of these changes in health care falls within the R&D development cycle of our current early-stage projects, we ought to prepare now.
On the biomarker side, NGS may contribute tools to study drugs as well as data that facilitate patient selection. Genomic information may sharpen our picture of drug action during short intervals of clinical and preclinical study; for instance, monitoring crucial mRNA populations in accessible tissues like blood. We seek earlier and more accurate indications of drug activity in the key aspects of target modulation, therapeutic efficacy, and safety. The relative ease of merging data, the sensitivity and the precision of RNA profiling via NGS (RNA-seq) confer an advantage over conventional microarrays.
Perhaps the best hope for short-term impact of NGS data lies in the capacity to discriminate drug-responsive patients from nonresponders, especially in oncology where somatic mutations drive many of the relevant traits. This task should be comparatively easy because we don’t necessarily need to unravel causation; association may be sufficient.
A handful of predictive genetic markers already exists in oncology, including Her-2/Neu and K-ras. Progress in other indications will likely require huge datasets that yield probabilistic, rather than categorically predictive, information. Setting aside patent issues, NGS may quickly displace other types of genetic tests when the error rates are reduced somewhat, simply because the cost of garnering an entire genome’s worth of allele information may be less than some of the current single-gene tests.
Another area where NGS may impact therapy in the short-term involves the assessment of acquired drug-resistance at the genetic level. NGS applications in oncology and infectious disease may enable physicians to tailor therapeutic regimens specifically to address recurrent disease. For human-scale genomes, we will likely need to reduce current NGS error rates to gain maximum benefit.
Target discovery remains our greatest challenge. As an industry, we are starved for novel disease mechanisms, for example, in the neurology area. In this realm it is vital to understand cause and effect.
When the first drafts of the human genome materialized a decade ago, rousing analogies like “human blueprint” and “periodic table of biology” blazed through the scientific and lay media.
Gradually a more levelheaded outlook emerged. The genome project yielded valuable infrastructure and a “parts list”, but hardly the recipe to reverse-engineer—let alone rebuild—a human being. Vital detail about gene function was lost in a wilderness of billions of bases and thousands of genes.
Nature’s gift of genetic variation has often provided a trail head amid confusing biological complexity. NGS machines can spit out human genomes worth of sequence, hundreds at a time, ultimately revealing the genetic diversity of the human species. The speed with which some disease-causing mutations can be defined with NGS technologies is eye-catching. About fifteen years ago, I was part of a large team that isolated the BRCA1 gene (captured beautifully in a book by the editor of Bio•IT World) after years of effort. The recent tour de force using NGS to pinpoint the gene involved in Miller’s syndrome contrasts sharply with the BRCA1 history in the time and number of personnel involved.
But will the sought-after targets for common grievous illnesses snap into focus with more, higher-resolution genetic studies? Countless genome-wide association studies (GWAS) have pointed to genetic determinants for many common diseases. But the vast majority of newly identified loci contribute weakly to overall risk. We lack the precision to infer meaning from most of these studies. At best, they recommend hypotheses; at worst, they may tell us little about the real pinch points for therapeutic intervention and send us on more fruitless drug discovery campaigns.
Optimists will remain optimistic, believing that the addition of one more log of statistical power or one more type of genomic data will unveil life’s mysteries. Pessimists will continue to point out that utility has lagged far behind data aggregation in the genomics era. The cryptographic simplicity of the genetic code and the elementary switch-like mechanism of the lac operon give a misleading impression of the complex disease biology in which we’d like to intervene. Even with the substantial effect on breast cancer risk conferred by BRCA1 mutations, the direct link to therapeutic target discovery remains tenuous.
It’s a thrill to glimpse the end-game for genetics—comprehensive genotype/phenotype correlations at the quantum level of genetics, the single nucleotide. No genetic footprint can remain hidden for long, given the remarkable technical feat of NGS.
But it’s also a bit scary. What if we can’t deduce the next generation of drug targets and mechanisms from the multiplicity of sequence data? Structural biologists can attest that, though atomic resolution crystal studies are a key asset in drug discovery, they haven’t obviated the need for a great deal of hard work and trial and error by smart medicinal chemists. •
Alexander (Sasha) Kamb is Head of Neurosciences at Amgen. He can be reached at firstname.lastname@example.org.