1. 1 Sequential Screening S. Stanley Young NISS HTS Workshop October 25, 2002

2. 2 What are the problems? HTS data sets are very large. Mixture of mechanisms. Bioassay data measured with error. Search space is very large.

3. 3 Statistical Analysis Corporate Collection Commercial Compounds Virtual Compounds Statistical Model Selected Compounds Initial Compounds Active Compounds Filters Screen Compounds

4. 4 Find Multiple Chemical Classes

5. 5 Separate Mixture into Components 10,000 x 20,000 Inactives

6. 6 MAO Inhibitors NN-CC C(3,1)-6-C(1,2)

7. 7 Individual versus Group (Compound versus Feature) Individual observations Distributions of means of 10 observations Distributions of means of 25 observations

8. 8 Small Shifts = Big Gains -40-20020406080100 N= 372 -40-20020406080100 N=23,232 Initial Screening Set Selected Compounds

9. 9 Enrichment Results Assay Training Pachinko Ratio #1 0.26 6.18 24x #2 0.6 0.3 Fail* #3 2.2 22.3 10x #4 2.1 15.0 7x #5 9.4 5.2 Fail* #6 2.6 42.5 16x #7 0.4 15.3 38x #8 0.16 1.5 9x * Assay not stable over time.

10. 10 Descriptors 1.Capture relevant chemical information. 2.Fast to compute. 3.Easy to interpret. 4.Not too many.

11. 11 Memorize versus Generalize Atoms / Functional Groups : -OH O(1,0) O(C ) HBD HBA =O O(1,1) O(C ) HBA Molecular twins in data sets.

12. 12 Definition of Terms Example: Pharmacophores for estrogen receptor (ER) –1D: Molecular Volume: 346.47 Å –2D: –3D: H-bond Donor H-bond Acceptor 7-10Å 2.5-4Å 10-14Å

13. 13 DE ABC Conformer 1 X Y Z RW Y SV Example of a conformer fingerprint ……….. Fingerprint Number 12345678 105471054810549 01111000001 PharmPrint: Theory C AB X Y Z D AC X Y Z D AB X Y Z C AB W Y Z D AC W Y Z E DB W Y R C DB W Y R Conformer 1

14. 14 Fragment Molecular Descriptors Atom pair N(3,0) - 7 - S(2,0) Atom triple N(3,0) - 7 - S(2,0) S(2,0) - 6 - N(2,0) N(3,0) - 12 - N(2,0) Topological torsion N(3,0)C(2,0)-1-C(3,1)O(2,0)

15. 15 Fragment Molecular Descriptors Augmented Atoms N(CCC), S(CC) Augmented Atom Pair N(CCC) - 6 - S(CC)

16. 16 Carry out Partial Least Squares Analysis (PLS) Use both active compounds and inactive compounds PharmPrint : Theory Cx-1 11100000000Molecule 3 01010000000Molecule 2 …..11111000001Molecule 1 11111100000 Molecule X 11011000000Molecule X-1 :::: :::: ….. 1 1 1 0 0 :::: 3-point FingerprintsActivitiesCoefficients C3 C2 C1 Cx :::: = X

17. 17 Example Catalyst Pharmacophore Features Aliphatic Hydrophobe Hydrogen bond acceptor Aromatic Hydrophobe Hetero-atom Projected location of hydrogen Aromatic Ring Hydrogen bond donor

18. 18 Model of AR Binding Site (Cool Picture) ARG752 GLN711 ASN705 2.94 Å 3.04 Å 2.79 Å 3.24 Å 10.50 Å