2. Overview
A little bit about why we make drugs, and
how computational chemistry is used (my
day job)
A little bit about confined electronic
systems, informational entropy, and
complexity (my evening job)
A novel 3D QM based structural descriptor
(my afternoon job?)
3. Drugs - why do we make them?
1. Money
2. And I guess you can help people too
3. But mainly for the money
4. Drugs - how do we make them?
From a computational perspective I will
limit myself to Structure-Based Drug
Design (SBDD)
5. Drugs - how do we make them?
What are we trying to do?
+ = ?
In SBDD we use computational chemistry to capture some
part of this incredibly complex interaction by modeling the
protein-ligand binding event
We typically ‘ignore’:
Protein flexibility, polarization and other electronic factors,
solvent, entropy . . .
6. And what have I been doing?
Detailed analysis of the in-house high-throughput virtual
screening protocol
Accepted in J. Chem. Inf. Mod.
Fragment-based de novo design
CONFIRM
Submitted to J. Comput.-Aided Mol. Design
A large scale critical assessment of docking programs
Binding mode prediction
Enrichment rates in virtual screening
Method development: Docking pose assessment tool
7. The Hospital that ate my Wife
Given the tools of our trade:
I can still work on problems in electronic structure
Information theoretic properties of strongly
correlated systems
Prof. Kalidas D. Sen, University of Hyderabad
Dr. Ali Alavi, University of Cambridge
8. Electrons and how they get along
PhD in small model quantum systems
Particles-in-a-box
Exact solutions
Archetypal systems for investigating electron
correlation
Electron correlation arises as a consequence of the
simultaneous interactions of mutually repelling
particles
It is what makes QM a ‘tricky’ problem both
conceptually, and practically
9. Basic physics of these systems
Two regions of behaviour
Small R - kinetic dominance
Large R - Coulombic dominance
E ~ A/R2 + B/R + …
Wigner ‘crystal’ formation at large R
10. Properties of interest
r r
Eigenvalues and
eigenvectors:
E i , "i (x1...x N )
r 2 r r
Density: n (r1 ) = N ! | " | ds1dx 2 ...dx N
Second order rr N(N - 1) 2 r r
density matrix: n 2 (r , r ') = ! |" | ds1ds 2dx 3 ...dx N
2
Physical rr 2 rr r
exchange- n xc (r , r ') = r n 2 (r , r ')# n (r ')
correlation n (r )
hole:
r r 2 r r
First order
density matrix:
$ 1 (x, x') = N ! | " | dx 2 ...dx N
FCI, RHF, UHF, and LDA solutions for both the spherical
(N=2, 3, 4, and 5) and cubic/planar (N=3, and 4) geometries
12. Spherical two electron system
RHF solution is surprisingly simple (S=0)
1 µ max
" (r) =
4#
$ µ =1
Cµ j 0 (% µ 0 r)
And rapidly convergent for even large R
! max=7)
(µ
13. Spherical two electron system:
RHF and informational entropy
Sr = " $ # (r) ln[ #(r)]dr
S p = " $ % (p) ln[% (p)]dp
ST = Sr + S p
!
16. A novel descriptor?
Doesn’t Sr look a little familiar?
Continuous form of a measure used in molecular
similarity:
S = "# pi ln[ pi ]
i
Could we use Sr as a measure of similarity?
Moreover, could Sr be a 3D QM-based structural
descriptor?
!
Literature search has shown that this has not been
considered before (I think)
17. A novel descriptor?
We want to make this useful
But we still have the problem of finding ρ in a timely fashion
Why don’t we approximate ρ?
We construct a pro-molecular density from a sum of fitted s-
Gaussians
"(r) # " Mol (r) = % "$ (r) = % % c$i exp(&'$i (r & R$ ) 2 )
$ $ i
Turns out that this isn’t as bad as you might think
!
18. Homebrew quantum mechanics
All of this has been done on my iMac at home
Molecular integrations performed using the
Becke/Lebedev grids in PyQuante[1]
Co-opted James into doing MathCad checks for
me. . .
[1] Python Quantum Chemistry - http://pyquante.sourceforge.net/
20. Homebrew quantum mechanics
Molecule Sr
H2O -7.42
H2S 3.94
Benzene -27.09
Cyclohexane (chair) -35.94
Perhaps Sr isn’t that discriminatory?
Plan B - Sr (r) = " #(r)ln[ # (r)]
22. Conclusions and outlook
Hopefully you have a feel for what I have been
working on, and why it might be interesting/useful
Work with Prof. Sen is being written up
Extend to planes - see if signature holds for N>2
At BI incorporate descriptor into a QSAR model
Is it of any use at all - what about Sp?