0.25
0.5
0.75
1.25
1.5
1.75
2
Virtual Screening and Library Design
Published on Jun 28, 201961 Views
Chemical space is estimated to contain ca 10^60 distinct small molecules; however in practice only tens of thousands (in more complex assays) to hundreds of millions (in DNA-encoded libraries) of comp
Related categories
Chapter list
Chemical Space (Relevant to Drug Discovery), Virtual Screening and Library Design00:00
Outline00:29
‘Drug-relevant’ chemical space02:21
Current drug and their targets03:21
Targets change over time05:04
Disease drivers and drug targets are quite different beasts!06:54
These were current drugs… how about ligandable proteins, those involved in disease?07:59
Current drugs focus on tiny amount of proteome (3%)09:29
Proteins with disease associations in OMIM (n=3,644) – MoA diversity depends on target class10:34
‘The rich get richer’… also current work (eg grants) is focused on few proteins11:45
So what is current ‘drugged’ and ‘druggable’ protein space?12:39
Aside: Do selective drugs cause fewer side effects? Not really…13:39
Are we looking for any particular physicochemical properties of drugs? 14:50
‘Drug-likeness’ is a very timedependent property!16:20
Number of new drug approvals shown to correlate with increase in molecular weight!17:09
Number of new drug approvals shown to correlate with increase in molecular weight! - 217:28
Currently drugged proteome and requirements for chemistry17:48
Outline19:09
Virtual Screening and Similarity Searching19:09
‘Similar Molecules Have Similar Properties’19:33
(Ligand-based) virtual screening20:22
Similarity Searching Requires an “Abstract” Representation (Descriptor)21:31
Similarity Searching in Practice22:09
The effect of sorting a database according to similarity to a query22:44
Numerical Performance Measure: Fraction of Actives in Top n% of Ranked Database23:00
Problem: Evaluation!23:22
Difficulties with validations (2)24:21
Descriptors: No ‘natural’ way to describe chemistry25:24
Descriptor Choice – What Is A ‘Tree’?26:18
1D/2D/3D (4D/5D/6D/…) Descriptor Classification27:25
Standard Descriptors – 1D28:22
Standard Descriptors – 2D29:12
Example of a well-performing 2D descriptor: “Circular Fingerprints”29:26
MDL keys… sometimes unintended uses work fine!29:36
3D Descriptors: Invariance w.r.t. translations, rotations and (sometimes) conformations is important30:46
Pharmacophoric Descriptors – Going 3D31:11
The Conformational Problem – Why 3D Can Be Tricky31:48
The Conformational Problem – Why 3D Can Be Tricky - 232:24
Representation is one step – comparison of molecules the next34:15
Comparing Molecules: Similarity Coefficients34:32
Similarity vs Distance Coefficients35:17
Is dissimilarity the opposite of similarity?36:45
The importance of shape – overall similarity can be very misleading!37:39
Similarity is not intrinsic property of objects38:10
Representation using biological information – HTS, docking, predicted targets, GE, images, …38:51
Is it possible and sensible to define “molecular similarity”?39:58
Library design, eg for high-content screens40:22
Chemical space is not equal41:02
Quite a number of different properties relevant for ‘drugs’42:45
Representation matters43:18
‘Maximum Diversity’ can give you unsuitable compounds!44:12
Problem with diversity… can we actually pick a ‘representative subset’?44:50
Practical aspects of diversity selection for screening45:57
Empirically frequently used solution46:37
Diversity library design conclusion47:57
Summary48:40