inhibitor design

Jie Liang jie at tgevax.life.uiuc.edu
Thu Feb 24 19:00:43 EST 1994


verlinde at u.washington.edu (Christophe Verlinde) writes:


>A big problem, however, with in computro screening is that our scoring
>functions are pretty poor. As a result many of the 'leads' found by
>these methods are inactive, some of the them are milimolar inhibitors,

I don't have experience in either protein-docking problem or  
x-crystallography, but it is such an important question that I very much 
would like to know the answer, too. 

Maybe I don't know the literature well enough, but I thought 
this raises another interesting question: why we have a poor scoring 
function?  

I can think of three reasons right now: 
1)  our knowledge of physicochemical interactions between atoms
of the molecules is primitive, all the electrostatic, free energy, etc.
computation do not improve things tremendously; 
2) the modeling of geometric complementation between molecules 
is unsatisfactory.  Usually to achieve realistic computation, some 
simplification of the geometry representation is necessary (such 
as the ball clusters and grooves etc in the DOCK program).  Some of the 
algorithms I have seen even allow some shape overlapping, etc. 
3) the rigid-body assumption really hurts and it needs to be removed, 
the dynamic shape matching hopefully should improve things.
"soft-docking" is probably too crude by introducing overlapping and
smoothing, etc.

My suspicion is that accurate geometric shape matching is the first
essential step: this is probably the most obvious path that nature hints 
us. If such obvious first order constraints are ignored, it is doubtful
that electrostatic and free energy calculations will buy us much.  

A recent paper by Lin and Nussinov et al. in _Proteins_ developed a set
of critical points to represent conciesely the protein surface.  I am
wondering, would adoption of such improved geometric representation
noticably improve the existing scoring function?  Is such algorithm
efficient enough for construction of those critical points and 
for real implementation of docking problem? 

Thank you for reading this.


Jie Liang

-- 
 Jie Liang ("Kommissar")         Internet:jie at tgevax.life.uiuc.edu
 156 Davenport Hall	                  liang at cs.uiuc.edu 
 607 S.Matthews Ave	  	"Working Hard Advancing Science to       
 Urbana, IL 61801                Corrupt the Mind." -- Black Uhuru  




More information about the Xtal-log mailing list