The rise and fall of therapeutic rationality

This ProPublica story—not just the spread of disinformation about these drugs, but specifically doctors’ complicity in generating runs and shortages, endangering patients who need them for chronic diseases such as lupus—reminds me of what the physician-historian Scott Podolsky calls a “pyrrhic victory” in the battle over “therapeutic rationality” in his wonderful book The Antibiotic Era: Reform, Resistance, and the Pursuit of a Rational Therapeutics—which anyone interested in the history or philosophy of medical evidence should go read immediately.

Podolsky shows that in the 1970s a powerful backlash from a coalition of doctors and pharmaceutical companies against the FDA’s new power to regulate drugs helped ensure we have no robust, centralized public oversight of prescription practices. (If you’re surprised to see doctors opposing what you think of as the public good, you’ll be even more surprised to read about their opposition to universal health insurance in Paul Starr’s The Social Transformation of American Medicine: The Rise of a Sovereign Profession and the Making of a Vast Industry.)

Here’s how Podolsky puts it:

The limits to government encroachment on the prescribing of antibiotics in the United States would be reached with Panalba and the fixed-dose combination antibiotics. While the FDA had been empowered to remove seemingly ‘irrational’ drugs from the marketplace, no one had been empowered to rein in the seemingly inappropriate prescribing of appropriate drugs. The 1970s would witness ongoing professional and government attention given to the increasingly quantified prevalence of ‘irrational’ antibiotic prescribing and its consequences, and such attention would in fact lead to attempts to restrain such prescribing through both educational and regulatory measures. The DESI process, though, had generated a vocal backlash against centralized attempts to further delimit individual antibiotic prescribing behavior in the United States, resulting in generally failed attempts to control prescribing at local, let alone regional or national, levels in the United States.

In the case of antibiotics, the result has been decades of promiscuous prescription, as overuse of antibacterials helped to breed a new generation of antibiotic-resistant “superbugs”—at the very same time that pharmaceutical companies, deciding that these drugs weren’t profitable, stopped trying to develop new ones. We thus have very few antibiotics to take the place of the ones that no longer work, even though isolated voices have been sounding the alarm all along—just as others have regarding pandemics. (Obama’s administration not only put in place a pandemic response team that Trump’s administration dismantled. It also developed a “National Action Plan for Combating Antibiotic-Resistance Bacteria.”) This is maybe the least familiar massive negative market externality of our time. Another result of such promiscuous prescription is much better known: we call it the opioid crisis.

However you view the FDA today—emblem of consumer protection or bureaucratic mismanagement, regulatory capture or government barrier to innovation, success story or failure—there is no question that public oversight of drugs is important and that it is high time to rethink how we regulate prescriptions, too.

Mechanism in biology

William Bechtel, Discovering Cell Mechanisms: The Creation of Modern Cell Biology (2006)

William Bechtel, “Mechanism and Biological Explanation,” Philosophy of Science (2011)

William Bechtel, “Biological Mechanisms: organized to maintain autonomy,” in Systems Biology: Philosophical Foundations (2007)

Carl Craver and James Tabery, “Mechanisms in Science,” Stanford Encyclopedia of Philosophy (2015)

Carl F. Craver and Lindley Darden, In Search of Mechanisms: Discoveries Across the Life Sciences (2013)

Margaret Gardel, “Moving beyond molecular mechanisms,” Journal of Cell Biology (2015)

Daniel J. Nicholson, “The concept of mechanism in biology,” Studies in History and Philosophy of Biological and Biomedical Sciences (2012)

Rob Phillips, “Musings on mechanism: quest for a quark theory of proteins?” Journal of the Federation of American Societies for Experimental Biology (2017)

James Tabery, Monika Piotrowska, and Lindley Daren, “Molecular Biology,” Stanford Encyclopedia of Philosophy (2015)

“White box” machine learning

From “A White-Box Machine Learning Approach for Revealing Antibiotic Mechanisms of Action,” Cell 177, 1–13 (2019):

Data-driven machine learning activities are poised to transform biological discovery and the treatment of human disease (Camacho et al., 2018, Wainberg et al., 2018, Webb, 2018, Yu et al., 2018a); however, existing techniques for extracting biological information from large datasets frequently encode relationships between perturbation and phenotype in opaque “black-boxes” that are mechanistically uninterpretable and, consequently, can only identify correlative as opposed to causal relationships (Ching et al., 2018). In natural systems, biological molecules are biochemically organized in networks of complex interactions underlying observable phenotypes; biological network models may therefore harbor the potential to provide mechanistic structure to machine learning activities, yielding transparent “white-box” causal insights (Camacho et al., 2018, Yu et al., 2018b).

Chemical and genetic screens are workhorses in modern drug discovery but frequently suffer from poor (1%–3%) hit rates (Roses, 2008). Such low hit rates often underpower the bioinformatic analyses used for causal inference because of limitations in biological information content. Experimentally validated network models possess the potential to expand the biological information content of sparse screening data; however, biological screening experiments are typically performed independently from network modeling activities, limiting subsequent analyses to either post hoc bioinformatic enrichment from screening hits or experimental validation of existing models. Therefore, there is a need to develop biological discovery approaches that integrate biochemical screens with network modeling and advanced data analysis techniques to enhance our understanding of complex drug mechanisms (Camacho et al., 2018, Wainberg et al., 2018, Xie et al., 2017). Here we develop one such approach and apply it to understanding antibiotic mechanisms of action.


Machine learning aims to generate predictive models from sets of training data; such activities are typically comprised of three parts: input data, output data, and the predictive model trained to compute output data from input data (Figure 1A; Camacho et al., 2018). Although modern machine learning methods can assemble high-fidelity input-output associations from training data, the functions comprising the resulting trained models often do not possess tangible biochemical analogs, rendering them mechanistically uninterpretable. Consequently, predictive models generated by such (black-box) machine learning activities are unable to provide direct mechanistic insights into how biological molecules are interacting to give rise to observed phenomena. To address this limitation, we developed a white-box machine learning approach, leveraging carefully curated biological network models to mechanistically link input and output data (Yu et al., 2018b).

h/t Anne Trafton of MIT News, “Painting a Fuller Picture of How Antibiotics Act”:

Markus Covert, an associate professor of bioengineering at Stanford University, says the study is an important step toward showing that machine learning can be used to uncover the biological mechanisms that link inputs and outputs.

“Biology, especially for medical applications, is all about mechanism,” says Covert, who was not involved in the research. “You want to find something that is druggable. For the typical biologist, it hasn’t been meaningful to find these kinds of links without knowing why the inputs and outputs are linked.”

An engine of discovery

from the preface to Cell Biology by the Numbers, Ron Milo and Rob Phillips:

One of the great traditions in biology’s more quantitative partner sciences such as chemistry and physics is the value placed on centralized, curated quantitative data. Whether thinking about the astronomical data that describes the motions of planets or the thermal and electrical conductivities of materials, the numbers themselves are a central part of the factual and conceptual backdrop for these fields.  Indeed, often the act of trying to explain why numbers have the values they do ends up being an engine of discovery.