A truth about Michael Levitt

Scott Le Grand legrand at teslaNOSPAM.mbi.uclaFNORD.edu
Thu Oct 2 23:21:36 EST 1997

Jerry Tsai wrote:
> Within the last month, I have come across some rather critical comments
> regarding Michael Levitt, and I think they are an unfair representation
> of Michael's accomplishments.  I did obtain my degree with Michael, and
> so I have my biases.  Mostly, I would like to offer another point of
> view, so that people who have not had contact with Michael can make a
> more informed judgement.
> In 1975, Michael published a paper in Nature that used simple potential
> to fold a reduced representation of a protein.  It has been said that
> this paper was not well thought out and hurt Michael's reputation.  The
> results were certainly controversial and from what I have heard, caused
> a great deal of discussion.  (In 1975, I was still in elementary
> school.)  Even though the paper was shown to contain certain weaknesses,
> it still has a great deal of merit and bearing upon current work.  Proof
> of this is that this paper has been cited at least 42 times in the past
> 3 years and over 500 times since it was published (data from SciSearch
> at LANL).  From what I have been able to put together, this publication
> did not seriosuly hurt Michael's reputation.  He has certainly not spent
> the last twenty years trying to rebuild it, nor has he had to learn any
> lesson from it.  In fact, this paper along with one published in JMB
> helped initially establish him in the protein modelling field.

Since I'm the guy who brought up Michael Levitt, I ought to reply here.

What's interesting about the two papers is that they were literally
waved (well at least cited :-)) in front of me as exactly what's wrong 
about trying to fold proteins by computer.  And this was in 1990, not
22 years ago mere weeks after I folded up an alpha helix for the very
with a genetic algorithm (though Mark Friedrichs beat us all by a good
2 years and never published anything).  Reading the two papers and
derisive comments about Dr. Levitt helped establish my skepticism of my

The two papers themselves are very creative and full of lots of good
but the bottom line is that they are a perfect example of losing one's 
objectivity when faced with promising results.  BUT, the problem in our
field is 
not that 22 years ago, a younger and more naive Michael Levitt made a
dumb mistake, but rather that too few people have learned anything since
then.  That's my central point, and that's what has to change.  You have
convinced me that a web page devoted to quoting claims of solving
aspects of 
the protein folding problem is long overdue.

Dr. Levitt is not alone in having made big mistakes early on in his
career.  Right
in this field, we also have Peter Kollman of UCSF, who was one of the
figures in the polywater flap, performing semi-empirical calculations
with Leland 
Allen in order to confirm the reality of the phenomenon. Once again, the
bottom line 
there was that the science was wrong, Dr. Kollman learned from the
experience, and he's 
made great contributions since then.

Finally, while it is silly to harp on someone's past, it's equally silly 
to deny it.  We all make mistakes.  It's the people that learn from them
that go 
on to make a difference.

> I also don't agree with the statement, "Don't be a Michael Levitt," as a
> warning to be careful about your work. In general, it carries certain
> negative connotations, and for people who don't know of Michael's work,
> does not truly represent his accomplishments.  Michael has been working
> on biological molecules for 30 years and his modelling work from RNA to
> proteins has been done carefully and is well respected.  Michael has his
> faults.  However, I don't think the context of the above statement's
> criticism is well founded.

It certainly does not represent his work since the 1975 papers,
but outside of the actual protein folders, his name and those papers
to come up whenever you start talking about solving the protein folding
problem by conformational search with experimental types.  Try it
with some 40-50something protein biochemists sometime.  

Scott Le Grand

More information about the Molmodel mailing list