1. Identification of structurally diverse Growth Hormone Secretagogue (GHS) agonists by virtual screening and structure-activity relationship analysis of 2-formylaminoacetamide derivatives. J. Med. Chem. 2004, 47, 4286-4290 GHS controls the release of growth hormone from the pituitary gland Therapeutically, GHS agonists replace direct growth hormone administration - growth deficiency - osteoporosis - obesity

2. Goal Identification of a new structural class or motif with activity comparable to that of known GHS agonists. Apply computational screening techniques to scan a medium-sized database of ~80000 compounds

3. Methods 2 screening approaches were used: Similarity searching -Fingerprints are generated for the query and database compounds (fingerprints are bit string representations of molecular descriptors) -Quantitative comparison using the Tanimoto coefficient (Tc) Tc = # of descriptors in common (intersection of query & target) total # of descriptors (union of query & target) Cell based partitioning -Chemical descriptors converted to low-dimensional representations, compounds partitioned into cells using binning schemes -Basic idea: compounds in the same cell are likely to display similar biological activity

4. Classification of molecular descriptors The authors used 2D fingerprints in their similarity search calculations.

5. Generation of low-dimensional representations For binning, the authors used systematic descriptor selection through a known algorithm to arrive at their cells

6. Results Similarity searching ~100 top scoring compounds were filtered (for Lipinski criteria) to get 58 compounds in the Tc = 0.6-0.8 range Cell based partitioning 61 compounds that fell into partitions containing T1 and T2 were selected Similarity Searching Cell-based partitioning Total 108 compounds identified 5 were active but… none of them were identified by both methods

7. Results 3 was the most active compound, came from similarity searching, 57% as active as T1

8. Refining the hit(s) Round one: analogue searching Conserve R1,R4 change R3’s: 117 compounds with varied R3’s tested Round two: synthetic chemistry Conserve R1 and optimum R3, change R2, R4: 18 compounds tested Conclusions: Diethylaminopropyl chain at R3 required. R4 has to be aromatic ring(s) R2 can be an alkyl chain (butyl). 3

9. Conclusion: Hit hopping achieved 43 Only 243 compounds out of ~80000 compounds had to be tested New hit potency is in the micromolar range (10-fold worse)

10. Discussion Binning algorithm Would 2D descriptor pharmacophore-based searching give better results? For larger databases or HTS, can you rank order hits after running the calculations/ virtual screens? How do you refine hits further?