Library Screening by Molecular Docking

Small molecule docking programs take a large database of molecules and screen each for complementarity to a binding site of known structure.  To distinguish this sort of docking from protein-protein docking (below), we will refer to it as “protein-ligand docking” or simply “ligand docking.”  To be useful, ligand docking screens must be fast, and therefore must make many approximations. Important energy terms, such as electronic polarization, are left out of docking calculations.  Important degrees of freedom, such as receptor conformation, are either under-sampled or simply ignored. In many ways, it’s surprising that docking screens work at all.

Docking has nevertheless had important recent successes^5-11.  The technique has discovered genuinely novel ligands for over 17 disparate targets in the last two years alone (Table 1).  Increasingly, docking predictions have been tested by subsequent determination of x-ray structures; this has particularly been true in the work of Klebe^10,12, Olson^13,14, and ourselves^15,16 (Figure 1).  Recently, docking screens have been compared to random high throughput screens (HTS), with the docking screens predicting ligands with a hit rate 100 to 1700-fold better than random screening alone (Table 2)^17,18.

How can the algorithmic weaknesses of docking be reconciled with these apparent successes?  Like any screening technique, docking tolerates both false positives and false negatives, as long as genuinely novel ligands are suggested at a rate high enough to justify the effort.  Its focus on libraries of available compounds makes it a popular technique for both pharmaceutical and academic screening¹⁹, and docking screens are now the most important way to leverage structure for novel ligand discovery.  As ever-more structures are determined²⁰, there is an ever-larger pool of potential users for the technique.

Unfortunately, docking screens are largely restricted to a small number of experts and their collaborators.  There are large barriers to entry into the field: small molecule databases suitable for docking are expensive to acquire, demand considerable curation, and the programs require expert knowledge.  Even those groups willing to purchase small molecule source databases such as the Available Chemicals Database (ACD) are frequently unprepared for the series of calculations, including assigning charges, solvation energies, and often conformations that are necessary to make the database useful for docking.  These barriers have diminished the impact of docking screens and limited their applicability.

Lowering these barriers to docking would bring the technology to a much larger audience.  To do so, the following would have to occur:

·         A large database of receptor structures must be accessible.

·         Binding sites on those receptors must be identified.

·         Large databases of compounds must be constructed, including:

¨       A library of purchasable small molecules (for experimental testing).

¨       A library of annotated drugs (for exploring possible receptor functions).

¨       A library of annotated metabolites (for exploring pathways and connections).

·         The interface to the docking software must be simplified.

·         Lower throughput but more reliable energy methods must be available for post processing.

·         This process must be available over the web and must be automated.

Read Also

Target	Best hit IC50 (mM)	Docking program	Structure solved?
Aldose reductase21	4.3	Adam & Eve	No
CDK422	44	Legend	Yes
Matriptase11	0.9	DOCK	No
Bcl-223	10.4	DOCK	No
Adenovirus protease24	3.1	EUDOC	No
AmpC15	26a	DOCK3.6	Yes
retinoic acid receptor25	2	ICM	No
TH receptor26	1.5	ICM	No
TGT10	8.3	LUDI/ FlexX	Yes
Carbonic anhydrase12	0.0008	FlexX	Yes
HPRTase27	2.2 a	DOCK3.6	No
Cavity site28	56 b	DOCK3.6	Yes
H2picolinate reductase18	7.2	FLOG	No
PTP-1B17	0.5	DOCK3.6	No
Edema Factor29	25 a	DOCK3.6	No
CK230	0.08	DOCK4	No

Table 1. Some recent docking successes (a, K_i. b, K_d.).

Technique	Compounds tested	Hits with IC50 < 100 μM	Hits with IC50 <   10 μM	Hit Rate (%)
HTS	400,000	85	6	0.021
Docking	365	127	18	34.8
Table 2. Docking versus screening for PTP-1B.