6.2. ilib diverse Filter Collection¶

The filter collection is grouped into five categories: default constraints, high drug-likeness, orally bio-available, blood-brain barrier permeable, and lead-likeness.

ilib diverse's default constraints¶

The default constraints set the compound's molecular weight between 100 and 700. The reactivity filter is enabled.

Filters for high drug-likeness¶

Ghose filter¶

More than 80% of the compounds of the Comprehensive Medicinal Chemistry database Ver. 97.1 were found with the filter developed by Ghose. The following drug-likeness constraints are applied to ilib diverse's Ghose filter:

MW: 160–480
Number of atoms: 20–70
logP: −0.4 to +5.6

Ghose, Arup K.; Viswanadhan, Vellarkad N.; Wendoloski, John J. A Knowledge-Based Approach in Designing Combinatorial or Medicinal Chemistry Libraries for Drug Discovery. 1. A Qualitative and Quantitative Characterization of Known Drug Databases. Journal of Combinatorial Chemistry (1999), 1(1), 55–68.

Lee filter¶

Lee developed a filter for high drug-likeness by the analysis of natural products and trade drugs. The following constraints are applied to ilib diverse's Lee drug-likeness filter:

MW: mean 356
logP: mean 2.1

Lee, M. L.; Schneider, G. Architecture and Pharmacophoric Properties of Natural Products and Trade drugs: Application in the Design of Natural Product-Based Combinatorial Frameworks. Journal of Combinatorial Chemistry (2001), 3, 284–289.

Mozziconacci filter¶

Mozziconacci developed a filter for drug-likeness by analyzing 15 commercially or freely available chemical libraries. The drug-likeness of these compounds was then investigated using common chemical features such as the Rule-of-5, the flexibility, the atom types and the functional groups. Based on this information, successive filters were designed to extract a drug-like subset of compounds. The following drug-likeness constraints are applied to ilib diverse's Mozziconacci filter:

RB: max. 15
Rings: max. 6
Oxygens: min. 1
Nitrogens: min. 1
Halogens: max. 7

Mozziconacci, J. C.; Arnoult, E.; Baurin, N.; Marot, C.; Morin-Allory, L. Preparation of a Molecular Database from a Set of 2 Million Compounds for Virtual Screening Applications: Gathering, Structural Analysis and Filtering. Institut de Chimie Organique et Analytique, Université d'Orléans.

Oprea filter (drug-likeness)¶

Oprea analyzed property distributions of physicochemical descriptors and properties in many databases containing drug-like compounds to identify optimal ranges. MDDR, Current Patents Fast-alert, CMC, Physician Desk Reference and New Chemical Entities were used as drug-like references and ACD as the non-drug-like data pool. The following drug-likeness constraints are applied to ilib diverse's Oprea filter:

HDO: 0–2
HAC: 2–9
RB: 2–8

Oprea, T. I. Property Distribution of Drug-Related Chemical Databases. Journal of Computer-Aided Molecular Design (2000), 14(3), 251–264.

Walters & Murcko filter¶

Walters and Murcko developed an extensive filter system for high drug-likeness. The following drug-likeness constraints are applied to ilib diverse's Walters & Murcko filter:

MW: 200–500
HDO: 0–5
HAC: 0–10
PSA: 0–120
RB: 0–8
Heavy atoms: 20–70
Charge: −2 to +2

Walters, W. Patrick; Murcko, Mark A. Library Filtering Systems and Prediction of Drug-Like Properties. Methods and Principles in Medicinal Chemistry (2000), 10, 15–32.

Filters for orally bio-available drugs¶

Fichert filter¶

The Fichert setting was developed using a Caco-2 model and 41 compounds. This cell model is widely used as an indicator of oral drug absorption. The following drug-likeness constraints are applied to ilib diverse's Fichert filter:

MW: max. 500
logD: 0–3

Fichert, Thomas; Yazdanian, Mehran; Proudfoot, John R. A Structure-Permeability Study of Small Drug-Like Molecules. Bioorganic & Medicinal Chemistry Letters (2003), 13(4), 719–722.

Lipinski filter¶

Lipinski's rule of 5 is the most approved filter for discrimination between drug-like and non-drug-like molecules. It was developed by analyzing the physicochemical properties of 2,245 compounds of the World Drug Index. Only compounds with INN (International Non-proprietary Name) or USAN (United States Adopted Name) are included in the study.

MW: max. 500
HAC: max. 10
HDO: max. 5
logP: max. 5

Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J. Experimental and Computational Approaches to Estimate Solubility and Permeability in Drug Discovery and Development Settings. Advanced Drug Delivery Reviews (1997), 23, 3–25.

Palm filters¶

These settings are based on the first study in which dynamic surface properties of drug molecules are correlated with drug absorption. The results suggest that the PSA (Polar Surface Area) is a better descriptor of intestinal drug absorption than log P.

Drugs with PSA > 139 Å² will be < 10% absorbed, while drugs with PSA < 63 Å² will be completely absorbed.

PSA max. 140 Å²: orally bioavailable
PSA max. 63 Å²: strictly orally bioavailable

Palm, Katrin; Luthman, Kristina; Ungell, Anna-Lena; Strandlund, Gert; Artursson, Per. Correlation of Drug Absorption with Molecular Surface Properties. Journal of Pharmaceutical Sciences (1996), 85(1), 32–39.

Palm, Katrin; Stenberg, Patric; Luthman, Kristina; Artursson, Per. Polar Molecular Surface Properties Predict the Intestinal Absorption of Drugs in Humans. Pharmaceutical Research (1997), 14(5), 568–571.

Veber filter¶

The Veber filter was developed by the analysis of 1,100 drug candidates. It seems that molecules fitting these two properties have a high probability of good oral bioavailability in the rat.

RB: max. 12
PSA: max. 140 Å²

Veber, Daniel F.; Johnson, Stephen R.; Cheng, Hung-Yuan; Smith, Brian R.; Ward, Keith W.; Kopple, Kenneth D. Molecular Properties That Influence the Oral Bioavailability of Drug Candidates. Journal of Medicinal Chemistry (2002), 45(12), 2615–2623.

Filters for blood-brain barrier permeable drugs¶

Murcko filter¶

Murcko used 1D and 2D descriptors based on Lipinski's Rule-of-5 and 2D fingerprints to determine CNS activity. The following constraints are applied to ilib diverse's Murcko filter:

MW: 200–450
logP: 0–5.2
HAC: max. 4
HDO: max. 3
RB: max. 7

Ajay; Bemis, Guy W.; Murcko, Mark A. Designing Libraries with CNS Activity. Journal of Medicinal Chemistry (1999), 42(24), 4942–4951.

Darvas, Ferenc; Keseru, Gyorgy; Papp, Aacytelkos; Dorman, Gyorgy; Urge, Laszlo; Krajcsi, Peter. In Silico and Ex Silico ADME Approaches for Drug Discovery. Current Topics in Medicinal Chemistry (2002), 2(12), 1287–1304.

Van de Waterbeemd filter¶

In contrast to drug-likeness, logP as a descriptor of the blood-brain barrier is not an effective classifier of CNS (Central Nervous System)-likeness. Van de Waterbeemd developed a filter for CNS-likeness with PSA (Polar Surface Area) and MW (Molecular Weight) as descriptors.

MW: < 450
PSA: < 90 Å²

Van de Waterbeemd, Han; Camenisch, Gian; Folkers, Gerd; Chretien, Jacques R.; Raevsky, Oleg A. Estimation of Blood-Brain Barrier Crossing of Drugs Using Molecular Size and Shape, and H-Bonding Descriptors. Journal of Drug Targeting (1998), 6(2), 151–165.

Filters for lead-likeness¶

Oprea filter (lead-likeness)¶

Oprea distinguished the lead-like chemical space from the drug-like space. The following lead-likeness constraints are applied to ilib diverse's Oprea lead-likeness filter:

MW: max. 450
logP: −3.5 to +4.5
HAC: max. 8
HDO: max. 5

Oprea, Tudor I.; Davis, Andrew M.; Teague, Simon J.; Leeson, Paul D. Is There a Difference between Leads and Drugs? A Historical Perspective. Journal of Chemical Information and Computer Sciences (2001), 41(5), 1308–1315.