Ligand Binding Site Prediction for HIV-1 Protease

1
Ligand Binding Site Prediction for HIV-1 Protease
using Shape Comparison Techniques
Manasi Jahagirdar1, Vivek K Jalahalli2, Sunil
Kumar1, A. Srinivas Reddy3, Xiaoyu Zhang4 and
Rajni Garg5 1Dept. of Electrical And Computer
Engineering, San Diego State University, CA,
2Dept. of Mathematics and Statistics, San Diego
State University, CA 3Molecular Modeling Group,
Indian Institute of Chemical Technology,
Hyderabad, India 4Chemistry and Biochemistry
Dept., California State University, San Marcos,
CA, 5Computer Science Dept., California State
University, San Marcos, CA
Introduction
Description
Discussion

• Effective binding site prediction is a primary
  step in the molecular recognition mechanism and
  function of a protein with an application in
  discovery of new HIV protease inhibitors that are
  active against mutant viruses
• Accuracy of binding-site prediction can be
  improved using a combination of shape descriptors
  for the interfaces
• We use geometrical, topological and functional
  descriptors in combination for ligand binding
  site prediction of HIV-1 protease

• The dataset for the algorithm for binding site
  prediction and extraction 90 HIV protease
  protein (21 wild type, and 69 mutated) PDBs
• The descriptors such as volume, dipole moment,
  moment of inertia, quadruple moment,
  hydrophobicity, residue interface propensity,
  integral of properties, and, Betti numbers are
  used for predicting the binding site
• The largest pocket of the protein is invariably
  the binding site for the ligand and hence residue
  interface propensity and hydrophobicity values
  are calculated for this pocket
• Predicted interface residues are residues with
  propensity gt 1.5. A propensity of 0 indicates
  that the amino acid has the same frequency in the
  interface and surface area
• For this dataset, ALA, ASP, ARG and VAL have high
  preference in the interface
• Predicted interface residues are distinctly
  hydrophobic.

Dataset Mutated and Wild Proteins
Residue Interface Propensity Values

• PDB 1B6J
• Mutation C67ABA, C95ABA, C167ABA, C195ABA
• 1B6J is a HIV protease complexed with macrocyclic
  peptidomimetic inhibitor

3D visualisation of protein, pocket and ligand
and descriptor information
Protein
Residue propensity and Hydrophobicity results for
protein pocket
Pocket
Future Work
Ligand

• Research and statistical results has proved the
  importance of utilizing a combination of
• descriptors in predicting binding sites of
  proteins. In the future, we plan to extend the
  algorithm to include more shape descriptors like
  tightness of fit, curvature in fine tuning the
  binding site prediction
• We plan to study the alternative sites for
  binding and the role of the attributes like
  volume, dipole moment, moment of inertia,
  quadruple moment, hydrophobicity, residue
  interface propensity, integral of properties,
  and, Betti numbers in the alternate binding site
  prediction
• This study can be extended for other HIV targets
  namely reverse transcriptase, integrase, gp41 and
  their inhibitors

Binding Site
Method

• Computational Approach
• Extract binding pockets present in mutated HIV
  protease proteins
• Assign various descriptors such as area, volume,
  inertia, electrostatic potential, Betti numbers,
  residue interface propensity and hydrophobicity
  to nodes in the pockets for matching score
  calculation and hence binding site prediction
• Residue Interface Propensity and Hydrophobicity
• Propensity for each amino acid is calculated as a
  fraction of the frequency that the amino acid
  contributes to the protein-ligand interface
  compared to the frequency that it contributes to
  the protein surface
• As per the scale we use, hydrophobic residues
  are Ala, Val, Leu, Ile, Pro, Met, Phe, Trp and
  Gly and the rest as hydrophilic

• Algorithm for extracting and
• comparing binding sites
• Compute a volumetric pocket function to represent
  the 3D shapes of protein pockets
• Compute an affine-invariant data structure called
  Multi-resolution contour tree (MACT) as a
  signature of the pocket function
• Compute and assign geometrical, topological and
  functional attributes to the MACT and check for
  compatibility of proteins and ligands by
  comparing their MACTs

References
1. Laskowski, R. A., Luscombe, N. M., Swindells,
M. B. Thornton, J. M. (1996) Protein Sci. 5,
2438-2452 2. Dong, Q., Wang, X., Lin, L., Guan.,
Y. (2007) BMC Bioinformatics, 8, 1471-2105 3.
Zhang, X. (2006) Volume Graphics 4. Campbell, S.
J., Gold, N. D., Jackson, R. M. Westhead, D. R.
(2003) Curr. Opin. Struct. Biol. 13,389395 5.
Binkowski, A., Naghibzadeg, S., Liang, J. (2003)
Nucleic Acid Research, 313352-3355 6. L.Young,
R.L.Jernigan, D.G.Covell. (1994) Protein Sci. 3
717-729 7. C.J.Tsai, S.L.Lin, H.J.Wolfson, R.
Nussinov. (1997) Protein Sci. 6 53-64

• Ligands are commonly found to bind with one of
  the strongest hydrophobic
• clusters on the surface of the target
  protein molecule
• If the distribution of residues occurring in
  the interface is compared with the
• distribution of residues occurring on the
  protein surface as a whole (residue
• interface propensity), a general indication
  of the hydrophobicity is obtained
• Combination of these two features appears to
  be a powerful tool for fine
• tuning the binding pocket surface area to be
  considered for binding site
• prediction of proteins