Sphere of Interest
Stanford Brain Lecture Notes
The RX Project:
CV Biblio (1985)
Index of Essays
Robotics, and AI
Health & Biotech
Earth Wisdom: Universe
Be Saved by Bob!!!
(And Other Balms )
Are Fats Killers
The Mystery of CONSCIOUSNESS
Who, What, When?
Review: Stan Dehaene's Consciousness & Brain
Near Death Experiences: In the Desert With Pim Van Lommel
Is the UNIVERSE
Fine-Tuned for Life?
Neuron Videos Say
Forget Realistic AI
EUV 2014 - Future of Moore's Law
BAM: Brain Activity Map of Spikes
What is Watson?
AI Overlord or Tool?
SETI: Search for Extraterrestrial Intelligence
KEPLER Seeks Earth-like Worlds
STEVE PINKER in the Amazon: photos
Billion Year Plan:
CONSCIOUSNESS as Global Resonance
Coronary Artery CT
Scan: A Life Saver
Book Review: TRANSCEND
Create a Mind
Does Drug X
Works in Monkeys!
Stanford Research: 1976-1986 Era
Automated Discovery: The RX Project
Imagine, as I did in 1976 when I came to Stanford, a future world in which every scrap of patient data was collected on computers: every symptom, every lab test or x-ray, every treatment. (That, of course, is the world we are in today.) In those days patient data was scribbled on scraps of paper bound into a manila folder.)
Then, imagine a computer tirelessly looking through that data trying to discover new relationships. What might it find? What are the causes of disease? What are the symptoms? What are the best treatments? Are there habits associated with good health and long life?
The computer would keep track of what it found and then try to verify it on other massive patient databases. It might announce its discoveries in the morning when the researchers gathered for their coffee. (Or, it might just undertake unauthorized research projects as in my Story of Ralph.)
That was the image behind my PhD thesis project at Stanford: the RX Project.
I designed and programmed the project from 1976 to 1981 and then, as Principal Investigator, I led a team of
students, statisticians, and computer scientists to refine and test it until 1986, when
I returned to clinical practice. (Now, I'm back on campus again, trying to integrate AI and
cognitive neuroscience - if such an integration is possible.) (BTW, my co-PI and
head of my PhD thesis committee was Prof. Gio Wiederhold, a database expert,
who at age 75 is still teaching in Stanford CSD.)
(Also on my thesis committee were
Profs Edward Feigenbaum (AI), Bruce Buchanan (AI),
Stanley N. Cohen
(Bioinformatics), Byron Wm Brown (Biostatistics), and James Fries (Med Informatics).
RX was an early example of data mining or exploratory data analysis under AI control.
It included a built-in knowledge base of clinical information. It also included
a “robot statistician” that it used to design its studies once a promising hypothesis
had been discovered.
Its discoveries would automatically be incorporated as new machine-readable knowledge.
The database that RX mined was the 1700 patient, multi-decade clinical database
ARAMIS, managed at Stanford by Prof. Fries for the American Rheumatism Association.
RX rediscovered several interesting and important causal relationships ( principally drug side-effects) and the system was widely discussed by medical computer scientists in the United States, Europe, and Japan. In 1985 I was among three scientists awarded the Toyobo Foundation's (Japan) top prize for bioinformatics.
Here are sections of the publication that most completely described this work -
Discovery, Confirmation, and Incorporation of Causal Relationships from a
Large Time-Oriented Clinical Data Base: the RX Project:
The RX Project - Abstract
Automated Discovery: the RX Project: (first 9 pages, 8mb)
(My most widely read paper on RX is from Computers and Biomedical Research in 1982. It is 24 pages long, so I've divided it into two pieces. This initial piece gives an overview.)
Automated Discovery: the RX Project (remaining 15 pages: 15 mb)
(The details and results are here for the truly motivated - longer download time.)
Below is a paper describing RX's most important discovery - prednisone elevates cholesterol.
It was published by the Annals of Internal Medicine in 1986.
Prednisone Elevates Cholesterol: A Discovery by RX (Abstract/page 1)
Prednisone Elevates Cholesterol (entire pdf: 14 mb)
(RX's most famous discovery. The Annals paper is a detailed description of the statistical
methods used to quantify this association in the face of missing and sporadic data.)
CAUSAL RELATIONSHIPS (CR's) are a crucial part of our knowledge of the world.
Here are some examples. Drinking too much alcohol can cause headaches. Exercise
promotes longevity. Burning fossil fuels increases atmospheric CO2.
The main task of RX was to learn CRs (ie to infer them) from the clinical data.
Once RX derived a causal relationship from its database, the next step was "learning it" by
incorporating the new CR into its knowledge base of clinical medicine. Its methods
are explained here "Representation of Empirically Derived Causal Relationships,"
"Modeling and Encoding Clinical Causal Relationships in a Knowledge Base."
Before starting work on machine discovery of causal relationships, I first worked
on the well known MYCIN Project, which automated infectious disease therapy.
In this task, MYCIN was shown by us to be comparable to infectious disease specialists.
Reports on RX by Gio Wiederhold et al. (RX related documents)
(This is a list of documents on RX and projects inspired by it by Prof. Gio Wiederhold.)
(Note: while the list is accurate, many links are unmaintained.)
The AI effort in RX included automating the tasks of the statistician
(selecting statistical methods and interpreting data sets) and the physician/epidemiologist
(choosing hypotheses, controlling confounders, and "publishing" (incorporating) results.
In RX's parlance, a causal relationship means a nonspurious, time-lagged correlation
between two variables. Nonspuriousness means that the correlation is not, in fact,
caused by some third (confounding) variable. In practice, this can be difficult to prove
in biological systems even with lab experiments and randomized, clinical trials.
RX's Discovery Module used time-lagged, nonparametric correlations
to detect initial hypotheses.
A key task of RX's Study Module was to use
the RX Knowledge Base
to attempt to control for known confounding variables,
however causally remote,
using directed acyclic graphs and multiple regression.