Title: Changes in the Profiles of Drug Properties: An Experimental, Computational and Informatics Perspecti
1Changes in the Profiles of Drug Properties An
Experimental, Computational and Informatics
Perspective
- Christopher A. Lipinski
- Pfizer Central Research
- Groton, CT 06340, USA
2Outline of This Talk
- ADME is rate determining rather than lead
finding - Experimental solubility in discovery and
development - Calculated trends in newer v.s. drug-like
compounds - Physico-chemical property trends depend on the
research approach, Merck v.s. Pfizer candidates - HTS hits, computational profile and experimental
solubility - Causes for computational trends at Pfizer, Groton
- Causes of poor aqueous solubility
- Effectiveness of computational filters
3Oral absorption problems (in-vitro to Candidate
stage) can be more costly in manning and less
predictable than the - to Lead and
optimization of Lead-in-Vitro stages
4Aqueous Solubility and Permeability Data Must be
Provided to Chemistry as Early as Possible to
Avoid Oral Absorption Problems
5Minimum Acceptable Solubility in ug/mL Bars shows
the minimum solubility for low, medium and high
permeability (Ka) at a clinical dose. The middle
3 bars are for a 1 mg/Kg dose. With medium
permeability you need 52 ug/mL solubility.
6The logic for a turbidimetric solubility assay
- Solubility in Discovery Solubility in
Development -
- turbidimetric solubility thermodynamic
solubility - non crystalline crystalline
- solids not characterized polymorphs
characterized - solubilized in DMSO solubility measured by
solid - added to stirred gavage equilibrating with
aqueous medium medium - 10s of minutes time scale 24 to 48 hours time
scale - used for early in-vivo SAR used for minimum
absorbable dose, dissolution, salt selection - correlation with in-vivo correlation with
clinical dosage animal SAR form - better in early discovery essential in
development
7Turbidimetric Solubility in a Flow Cell. lt 20
ug/mL is poor solubility.
- Used to measure solubility at 5-65 ug/mL for
poorly soluble heterocyclic compounds
8Computationally comparing libraries. Drug-like
v.s. new drugs
- Use the presence of an INN name or a USAN name or
marketed status as a flag for a compound with
drug-like properties - 7483 Drugs with INN name, USAN name or approved
for marketing - Compare to 2679 New Drugs from the Derwent World
Drug Index - mechanism field - trial preparations
- No CAS registry number, no INN/USAN name,
abstracted in 1997, 1998, 1999
9Distribution Parameters for 7483 INN/USAN Drugs
Define the 90 Limits Corresponding to Properties
Unfavorable for Oral Drug Absorption.
10The Rule of Five mnemonic.
- Poor absorption or permeation are more likely
when there are - More than 5 H-bond donors.
- The MWT is over 500.
- The CLogP is over 5 (or MLogP is over 4.15).
- The sum of Ns and Os is over 10.
- Substrates for transporters and natural products
are exceptions.
11How Pfizer Uses the Rule of 5
- On-line alert at compound registration
- Filter for HTS screening library
- Filter for purchased compounds
- Criteria for focused library synthesis
- Candidate nomination guideline
12Newer drugs (blue) are higher molecular weight
than drug-like compounds (green).
Molecular Weight
2679 Newer drugs, 7484 Drug-like compounds
13Newer drugs (blue) are less permeable than
drug-like compounds (green) as calculated with
polar surface area.
Polar Surface Area
2679 Newer drugs, 7484 Drug-like compounds
14Newer drugs (blue) have more rotatable bonds than
drug-like compounds (green).
Rotatable Bonds
2679 Newer drugs, 7484 Drug-like compounds
15Newer drugs (blue) have higher Andrews Binding
Energy than drug-like compounds (green).
Andrews Binding Energy
2679 Newer drugs, 7484 Drug-like compounds
16Newer drugs (blue) are slightly more lipophilic
than drug-like compounds (green) using the
Moriguchi LogP.
Moriguchi LogP
2679 Newer drugs, 7484 Drug-like compounds
17Upwards molecular weight trend in Merck advanced
candidates.
18Upwards molecular weight trend in Pfizer, Groton
early candidates
19No increase in lipophilicity with time in Merck
advanced candidates.
20Upwards lipophilicity trend with time in Pfizer,
Groton early candidates.
21Increasing hydrogen bond acceptor trend with time
in Merck candidates
22No hydrogen bond acceptor trend with time in
Pfizer, Groton candidates.
23Lead Approach, Permeability and Solubility
- Structure Based
- MWT up
- peptidomimetic
- fit 3 or more sites
- H-Bonding up
- fit _H bond sites
- fit salt bridges
- LogP no change
- no selection pressure
- Poor Permeablity
- HTS Based
- MWT up
- HTS selects for larger size
- LogP up
- HTS selects for high LogP
- H-bonding no change
- no selection pressure
- Poor Solubility
24Leads at Pfizer changed towards higher molecular
weight and lipophilicity with the advent of high
throughput screening.
25Early high throughput screening hits (blue) were
more lipophilic than phase-2 (green) or marketed
drugs (red).
26Filtering improved computational HTS hit
profiles, but solubility of HTS hits could still
be low (data from 633 Pfizer HTS hits)
2778.5 Percent of phase II compounds (INN, USAN or
NCE) have good aqueous solubility (gt 65 ug/mL).
Only 14.2 percent have poor solubility (data
based on 1597 solubility assays on phase II
compounds)
2831.2 Percent of 2246 commercially purchased
compounds have very low aqueous solubility (lt 20
ug/mL), incompatible with appreciable oral
activity
2939.8 Percent of 33093 medicinal chemistry
compounds have very low aqueous solubility (lt 20
ug/mL).
30Summing Up - Understanding the solubility problem
- Traditional Drug Discovery
- tight link to oral activity testing
- 10-15 poor solubility
- Single Compound Traditional Synthesis
- no link to oral activity testing
- isolated by crystallization
- 30 poor solubility
- Single Compound Lead Optimization
- traditional in-vitro optimization
- additional 10 poor solubility
- Combinatorial Chemistry ??
31Effectiveness of our computational intervention
strategy is measured by the decrease in our Rule
of 5 alerts
32Summing Up, Pfizers Experimental, Computational
and Informatics Perspective
- Drug property profiles are best improved by a
combination of - computational filtering initiatives
- discovery focused experimental assays
- cross departmental, people based, educational
initiatives - data base mining approaches