Big Knowledge (the next thing after Big Data) approach may offer an answer.

An early result of the COVID-19 treatment trial RECOVERY two days ago reveals that dexamethasone (a widely used corticosteroid) reduces death of patients on mechanical ventilators by one third, whereas it had no beneficial effect of patients who do not require respiratory assistance. Why does dexamethasone help? After all, corticosteroids treatment for COVID-19 has been dis-recommended by WHO, CDC and many experts.

We used a new computational approach, Big Knowledge (the next more intelligent thing after current brute-force Big Data) to look for molecular mechanisms by which the COVID-19 virus, SARS-Cov-2, destroys the lung tissue. Big Knowledge uses vast networks of millions of interconnected biomedical facts to allow researchers to automatically “connect the dots” within the totality of human biomedical knowledge –while ignoring discipline boundaries.

With the help of SPOKE, one of the largest biomedical knowledge networks for this new genre of research, developed by Sergio Baranzini at UCSF, we predicted two months ago (April 2020) the following: Dexamethasone could be effective in mitigating the pulmonary pathology specifically in severely ill COVID-19 patients who have been placed on mechanical ventilation. A brief report of the analysis can be found on a preprint server here.

Mechanical ventilation, while providing critical (life-saving) respiratory support, also stresses the lung tissue. Such cell-stress in turn upregulates the expression of the protein molecule ACE2 on cells — which ironically is also the molecular docking site that allows the COVID-19 virus to enter the cells and multiply, and then, infect other cells!

So obviously, we have a vicious cycle –the engine that drives all kinds of acute disease and precipitous death in medicine. The more the lung is damaged by the virus, the more you need to ventilate, the more you hence stress the lung tissue, the more you thereby upregulate ACE2, which in turn facilitates virus infection new cells — causing more damage…

By connecting the dots in the knowledge network, SPOKE told us that dexamethasone could specifically block the molecular pathway by which tissue injury upregulates the expression of the ACE2 gene. Thus, dexamethasone, or corticosteroids in general, would disrupt the vicious cycle.

The RECOVERY trial results now seem to validate our predictions. It is not hard to predict the possible utility of corticosteroids for any disease. But what is remarkable was that the Big Knowledge approach with SPOKE predicted that corticosteroids would differentially be of therapeutic benefit to patients receiving mechanical ventilation — as the trial now also shows. This granularity of a prediction suggests that relevant biology has been captured.

No medical journal wanted to consider our computational analysis results for publication. Not even the preprint servers bioRxiv and medRxiv. (Hence our manuscript ended up on the more open-minded European preprints.org)! The reason for the rejection can be suspected as being the following: (1) every time you propose an idea with an iatrogenic element in the cause of a disease, the establishment balks — as Ignaz Semmelweis would know. How sad that even bioRxiv and medRxiv are now part of the rigid system of censoring scientific thought by editors of scientific journals. (2) We arrived at our results in a novel, unfamiliar fashion (and editors don’t like new ideas that they cannot fathom!): use of a new genre of machine-dependent reasoning for connecting the dots to come up with a plausible testable hypothesis.

In fact, the testing of the hypothesis itself, by carefully monitored explorative treatment with corticosteroids, could already have saved lives of ventilated COVID-19 patients. Some critical care clinicians in fact have suspected that corticosteroids may help and had used them, producing anecdotal success cases, but they were silenced by the medical Establishment.

Medical progress and the veracity of medical facts are built on two pillars: (a) empirical data — and (b) logical connection of the data points. The latter is often bubbling up only as vague intuition in the mind of experienced clinicians. But in the future, with the arrival of Big Knowledge (that will complement Big Data) and the help of biomedical knowledge networks, such intuition can be placed on a solid mechanistic foundation.

Institute for Systems Biology