Universit degli Studi di Milano Dipartimento di Scienze

Università degli Studi di Milano Dipartimento di Scienze Farmaceutiche “Pietro Pratesi” Gri. Dock: An MPI-based software for virtual screening in drug discovery Alessandro Pedretti

What is the virtual screening ? • The virtual screening (VS) is a computational approach that can be used in drug discovery processes to find new hit compounds. Database of molecules Set of molecules Database filter Experimental assay Hit compounds Virtual screening High-throughput screening • It can be compared to the High-throughput screening (HTS) that is a true experimental approach.

The database of molecules • The database must contain molecules that are available in the real world or synthetically accessible in easy way. • The pharmaceutical industries have got databases built trough the years from researches in some different fields. • Some databases are publicly available and provided by chemical compound resellers (AKos, Asinex, Tim. Tec, etc) or by non-profit institutions (Kyoto University, NCI, University of Padua, etc). • The database must contain a large number of molecules in order to do an exhaustive exploration of the chemical space.

The database filter • The database filter does the virtual test to check if a molecule could be bioactive or not. • The kind of filter allows to classify the virtual screening approaches in: Ligand-based The 3 D structure of the biological target is unknown and a set of geometric rules and/or physical-chemical properties (pharmacophore model) obtained by QSAR studies are used to screen the database. Structure-based It involves molecular docking calculations between each molecule to test and the biological target (usually a protein). To evaluate the affinity a scoring function is applied. The 3 D structure of the target must be known.

Molecular docking + Ligand Receptor Docking software • The complex quality is evaluated by the score. Ligand – receptor complex

Gri. Dock – Main features • Gri. Dock is a software developed to perform structure-based virtual screenings. • It’s a front-end to the well known Auto. Dock software, developed by D. S. Goodsel and A. J. Olson. • It uses VEGA command-line software to perform file format conversion, database extraction and molecular property calculations. Auto. Dock 4 + VEGA Virtual screening Gri. Dock • Highly portable C++ code (Linux 32 and 64 bit, Windows 32 and 64 bit). • It can take full advantages of multi-CPUs/cores systems and GRID-based architectures through its parallel design.

How Gri. Dock works Database of molecules VEGA Receptor coord. + maps Auto. Dock 4 Ligand – receptor complexes Score analysis Output files • Calculation of the molecular properties. • Input file generation (PDBQT). • Molecular docking. • Score calculation.

How VEGA works with Gri. Dock Database of molecules Hydrogens add Property calculation AMBER force field Gasteiger-Marsili method Potential attribution Calculation of charges Search of flexible torsions Conversion to PDBQT to Auto. Dock 4

Gri. Dock multi-threaded version Database Receptor Thread loop Gri. Dock main thread Thread 1 Thread 2 Thread n VEGA Auto. Dock 4 Symmetric multiprocessing (SMP) provided by pthread library or Windows APIs Mutex controlled access Output files* • Log file (gridock_DATE. log). • CSV file containing the list of complexes ranked by docking score. • Zip file containing the output complexes generated by Auto. Dock 4.

Gri. Dock MPI version Gri. Dock MPI master node Node loop Database Receptor Node 1 Database Receptor Node 2 Database Receptor Node n VEGA Auto. Dock 4 Gri. Dock MPI master node Output files MPI

Gri. Dock input requirements To perform a virtual screening with Gri. Dock, you need: • The 3 D structure of the biological target. - Protein Data Bank (http: //www. rcsb. org). - Homology modeling. • The 3 D maps of the active site generated by Auto. Grid 4 - Auto. Dock. Tools / MGLTools (http: //mgltools. scripps. edu). - VEGA ZZ (http: //www. vegazz. net). • One or more databases of 3 D structures in SDF or Zip format. • Ligand. Info: Small-Molecule Meta-Database (http: //ligand. info). • MMs. INC (http: //mms. dsfarm. unipd. it/MMs. INC. html). • ZINC (http: //zinc. docking. org).

The Citrus tristeza virus case • The Citrus tristeza virus (CTV) is a positive single stranded RNA virus that causes a serious pathology of the citruses. • Any treatment to save the infected plants is unknown. • A possible therapeutic target could be the RNA-dependent-RNA polymerase (Rd. Rp) involved in the virus replication. Infected cell Protease ss. RNA (+) – 5’ prot. m. RNA prot. Translation Virions Translation Structural proteins Early protein Other proteins Protease Rd. Rp (-)RNA Replicative complex

The Rd. Rp model The crystal structure doesn’t exist and a homology modeling procedure was performed: Swiss. Prot Q 2 XP 15 Primary structure Fugue Folding prediction Rough 3 D structure Rd. Rp model VEGA ZZ + NAMD To the refinement workflow

Model refinement Rough model Missing residues VEGA ZZ + NAMD Side chains add Hydrogens add 30. 000 steps conjugate gradients Energy minimization Structure check Model ready for the screening Ramachandran plot

Calculation of the grid maps Auto. Dock requires pre-calculated grid maps to evaluate the total interaction energy between the ligand the target macromolecule. To do it, we used the script included in the VEGA ZZ package: Rd. Rp structure Potential attribution Calculation of charges Apolar hydrogens remove PDBQT file Mapping the active site Script file: Auto. Dock/Receptor. c Auto. Grid 4 run Grid map files

Considered databases All test databases in SDF format were downloaded from http: //ligand. info: • Chem. Bank • Chem. PDB • KEGG Ligand • Anti-HIV NCI • Drug/likeness NCI • Not annotate NCI • AKos Gmb. H • Asinex Ltd. The total number of docked ligands is: ~1, 000

Test system Tyan Transport VX 50 • # 8 AMD Opteron 875 dual core CPUs @ 2. 4 GHz. • 8 Gb Ram. • 72 + 150 Gb SATA hard disk. • Linux 64 bit (Cent. OS 4). 40, 000 ligands/day.

Preliminary results The top ranked ligands contains in their structure one or more sulfurs. Sulfonic acid derivatives. These compounds are know to be potent inhibitors of the HIV reverse transcriptase. Some of them are naphtalen polysulfonic acids developed as Anti-HIV (Anti-HIV NCI database).

Conclusions • We developed a new parallel structure-based virtual screening software able to run on both multi-CPU and GRID systems. • The complete model of the RNA-dependent-RNA-polymerase of Citrus Tristeza Virus was obtained to perform a virtual screening study. • Screening ~1, 000 ligands, potential Rd. Rp inhibitors were found. • These molecules contains sulfur atoms and, more in details, multiple sulfonic acid moieties. • Some of them are included in the Anti-HIV class. • To complete the study, the activity of the found molecules must be experimentally confirmed by biological assays.

Acknowledgments • Giulio Vistoli • Santo Motta • Cristina Marconi • Francesco Pappalardo • Alessandro Lombardo • Emilio Mastriani www. vegazz. net www. ddl. unimi. it