Elementary, But It’s the Wrong
Watson
By
Leonard Zwelling
I received an email from Houston
while in Durham, NC. I was back at Duke for my 40th medical school
reunion and thus did not see the article in Friday’s Chronicle reporting on the
scheduled news conference at Anderson to announce the new partnership between
the Anderson astronauts of the moon shot persuasion and IBM.
If I
understood what I read, the proposed research, essentially comparative
effectiveness research on steroids, will use data from one million leukemia
patients treated currently or in the past to ascertain trends or anomalies not
detected by our clinical investigators already. The new partner here is the
analytic power of a computer called Watson that beat human contestants in a
short answer speed test on Jeopardy. Hmmmmm……
The proposal is to feed Watson a huge amount of clinical and
clinical research data from Anderson leukemia patients and to use the machine’s
amazing power to detect findings heretofore missed.
Let’s sit back and think about the premises that must have
been employed to justify this exercise and expense:
All of the data
currently housed in the Anderson data bases will, once analyzed by Watson,
reveal insights into the nature of leukemia and of its optimal treatment.
Possible.
The cost of this
project is worth diverting funds from supporting other research for surely the
help of Watson has a large price tag (one estimate is $15M) that is the
opportunity cost associated with IBM’s help. Possible.
The likely
ultimate result will be the recognition of various sub-categories of leukemia,
some more amenable than others to past and current therapy. As with all such
big science projects, whether the data being analyzed are molecular or
clinical, this is likely to result in further disease category splitting unless
and until a broadly effective therapy is developed (e.g., the drugs used against
acute leukemia of childhood, testicular cancer or Hodgkin’s Disease). If a good
regimen is discovered that produces a 90% remission rate, the only
characterization of interest is that of the 10% of patients who don’t respond
at all. The need to split is generated by the relative ineffectiveness of
current drugs. Develop something that works, even if through old-fashioned
empiricism without a clear understanding of how it works, and this splitting
becomes far less important. Definitely.
It is not
necessarily the case that any new insights into the cancer biology of this
group of diseases will be acquired by
rediving into the same pile of data only faster and deeper. Who knows?
As
I have written many times in the past, I am not a fan of the reductionist view
of cancer as a disease into which we will gain insight by grinding it up into
tiny parts, contemplating the parts, and then putting them all back together. (A
friend describes this as trying to discern the identity of a type of car a pile
of iron filings used to be).
Did
we understand the origins of Acute Lymphocytic Leukemia all that much better
than the origins of AML when the cure of ALL was discovered? I think not, but
perhaps I am incorrect. However, the survival rates sure are different for ALL
vs. AML now.
Thus,
it is natural for me, a skeptic from the word go, to say we have brought in a
Watson as a consultant, but the wrong one.
IBM’s
Watson will work as hard as possible using technology built to do just that,
work we humans cannot do as quickly or efficiently. But I prefer the other
Watson’s approach.
The
other Watson is James Watson who was described by fellow Nobel laureate Max Perutz
as never confusing hard work with hard thinking. And real thinking, as I was
privileged to hear described this Friday morning at Duke Grand Rounds by Ralph
Snyderman, Chancellor Emeritus of Duke Health System, when he spoke of the life
of his close friend and the first Duke Nobelist Bob Lefkowitz (Chemistry 2012),
is done only by real people, not machines.
Solving
the cancer problem will take far more thinking than work for right now for we
really don’t know what work is likely to be the most productive. Until we do, a
combination of humility as well as perspiration and contemplation might be a
better approach than asking;
I’ll
take Cancer for $1000 Alex:
Answer:
They
spent more money in less time resulting in less impact than ever before
Question:
What
is MD Anderson?
In fact, that last one is a good question
for Dr. DePinho as well as for his new sidekick Watson—the non-Jim, non-Nobel
Prize variety.
