You are here

How to Rescue the Life Sciences from Technological Torpor

Having spent my career in two fields grounded in the physical sciences that made better, faster, cheaper a core driving principle—telecommunications and semiconductors—it’s hard not to cast a jaundiced eye at the sorry state of the pharmaceutical industry. To paraphrase the immortal Dean Wormer in Animal House, ineffective, slower, and more expensive is no way to go through life!

With the exception of DNA sequencing, which has enjoyed Moore’s Law-like improvements for a decade, drug discovery and development has failed miserably when it comes to harnessing a virtuous circle of ever increasing effectiveness and efficiency. In fact, it is moving in the opposite direction. This does not bode well for the future of a business under assault on so many fronts. Rearranging the deck chairs through waves of mergers and layoffs may temporarily fatten shareholder returns and executive bonuses, but such financial engineering will not cure cancer.

Real engineering is called for, but the pushback is enormous. The senior scientists and executives now running pharma often say that biological systems are too complex to decompose, model, emulate, and control using the systematic analysis, design, and development methodologies that propel the world of electronics. Perhaps that was true back when those industry leaders were first trained. But technology has taken great leaps forward, while most of them seem mired in the past.

It is precisely the success of modern electronics and computing that has made it possible to design, manage and operate the kinds of advance tools, simulation and automation techniques that will be needed to shift the life sciences over to practices that have been advancing the physical sciences for years.

Changing Culture

In order to do this the culture must change. Mathematics is the language of engineering and life scientists can no longer take a pass on it. A system that cannot be modeled cannot be understood, and hence cannot be controlled. Statistical modeling is not enough, for the simple reason that correlation is not causation. Life science engineers need to catch up with their peers in the physical sciences when it comes to developing abstract mathematical representations of the systems they are studying. Progress comes from constantly refining these models through ever more detailed measurements. R&D should be all about finding ways to couple the two.

Forty years ago, when I entered MIT, a great division took place my freshman year. Those who could handle the math stuck with engineering or the physical sciences, while those who couldn’t found their way into biology, chemistry, or medicine. The former imbued a language of precision, a respect for systematic problem solving, and a penchant for measurement and reproducibility. Meanwhile, the latter entered a phenomenological jungle where nomenclature reigned supreme, intuition substituted for rigor, and guess-and-check methodologies guided experimental investigation. This classic approach may have been useful for harvesting low-hanging fruit, but the thinning drug pipeline demonstrates that this strategy is running out of gas.

Sure, enormous amounts of data are required to get to the root of how living machines function, and collecting this Big Data requires Big Automation. Yet there is a stupefying amount of work that is still done by hand in most life science labs. This is enabled by the chronic glut of grad students, Ph.D.s, and post docs—an oversupply driven by public policy and federal spending that has allowed researchers to stave off automation.

The title of a recent news story in the Washington Post says it all:"U.S. pushes for more scientists, but the jobs aren’t there.” “Despite $10 billion in federal stimulus funds funneled through the NIH to “create or save” 50,000 science jobs, Ph.D. chemists and biologists are increasingly going begging.

A recent National Science Foundation survey tells the sorry tale. Three to five years after graduation, biological and life science doctorates are dead last in finding tenured or tenure-track faculty positions. Instead, many become part of what one economist calls a “pyramid scheme that enriches — in prestige, scientific publications and federal grant dollars — a few senior scientists at the expense of a large pool of young, cheap ones.”

Running this kind of medieval guild system at taxpayer expense not only dulls the minds of young scientists by forcing them to spend far too much time doing repetitive tasks, it pollutes the discovery and design process with non-systematic errors, confirmation bias, and the indiscriminate discarding of data from “failed experiments.” Is it any surprise that (at least according to one pharma company) two-thirds of the experimental results published in peer-reviewed life science journals cannot be reproduced?

The key to making Moore’s Law work is to scale everything. The popular press focuses on the ever shrinking physical dimensions of transistors, but the entire semiconductor ecosystem had to scale to keep pace, including the way it conducted science. This was made possible by the relentless detection and expunging of systematic error, which is only produced by machines. Human labor is not systematic, and therefore cannot scale. That brings us to the final piece of the puzzle.

The sophistication and reliability of the robotic systems that enable modern semiconductor plants to operate with atomic precision needs to be brought to bear on the woefully inadequate fluid handling, data collection, and assay technologies that beset the life sciences. And not just in high throughput screening but in the balance of research, all the way through to manufacturing.

Advancing the frontier of semiconductor research necessarily leveraged the same equipment and tools used in the factories. To this day, test wafers that form the platform for science experiments work their way through the semiconductor process flow right alongside product. A vast technology gap between R&D and manufacturing cannot be sustained.

Yes, Big Pharma uses sophisticated robotics to screen millions of compounds for potential hits. And then? They hand promising leads off to an army of medicinal chemists who labor by hand to turn them into drugs. That’s like driving a Porsche from New York headed for Boston just to get off in Hartford to take a horse the rest of the way.

Thankfully, these lessons are not lost on everyone. A new generation of computational biologists cross-trained in the engineering disciplines, now in their late twenties and early thirties, is beginning to make its impact on the industry. From my perch as an angel investor in a new-era pharmaceutical startup, still in stealth mode, I am beginning to see what happens when you turn these young people loose, unencumbered by oldthink.

Stand back and watch them rescue a business long overdue for change.

Bill Frezza is a Fellow at the Competitive Enterprise Institute and a Boston-based venture capitalist. He can be reached at If you would like to subscribe to his weekly column drop a note to

This article was originally published in Bill Frezza's column, Skeptical Outsider, at