BioByte logo left column
contact us


[standing baby]

Much like a baby just learning to stand, Drug Design using QSAR very often gets off to a shaky start. You become excited when a regression equation you are developing shows robust statistics (e.g. r2 > 0.9). Your 'Baby' looks so promising (even if somewhat bare-assed)!

Before popping the champagne cork, however, there are a few things you should check out:

  • Are the terms in your equations chemically meaningful?
  • Do they suggest the direction to take in synthesis of more effective analogs?
  • Do the parameters suggest a mechanism?
And before persuading your management to make a large investment in synthesis and testing of new compounds, based on your Baby's first efforts to stand, it makes sense to see if other research supports it. Why not take advantage of the power of Lateral Validation which is available in the database of over 14,000 equations (many of them unpublished) that is part of the C-QSAR package developed by Prof. Corwin Hansch (honored in 1990 as the "Father of QSAR" by the QSAR Society) and offered by BioByte Corp.

The paper by Hansch and Fujita1 which first proposed using Hammett-type parameters in biological QSAR quickly became a 'Citation Classic' and the foundation for a world-wide research effort. Chemical Reviews2 recently published their 50 most frequently cited articles in the last 50 years, and Jaffe's paper on the Hammett equation topped the list, followed by the paper by Leo, Hansch and Elkins on Partition Coefficients, illustrating how the octanol/water system has become the standard model for the hydrophobic parameter in biological QSAR. Instead of slacking off over the past 35 years, the pace of this research has quickened. In the past ten years the papers of Hansch and Leo, in the area of QSAR, have been cited over 100,000 times.

It is easy to find examples of published QSARs which meet acceptable statistical standards but which fail to guide their authors to any useful advances. When probed for chemical and physical meaning, they totter and fall flat. One such QSAR, widely advertised and included in a software vendor's User Manual3, uses data published by Kuipers et al4 on the binding of flesinoxan analogs at the dopamine (D2) and 5-hydroxytriptamine receptors (5-HT-1a). If the Kuipers group had the C-QSAR database at the start of their research, they would have quickly found 90 structures with dopamine activity (agonist or antagonist), many with measured hydrophobic parameters (log P oct) and all with values calculated by CLOGP. Looking at the regression equations for this activity type, they would have seen two for D-2 on rat striatum which were best fit with a bi-linear relationship in log P. The data for two equations on binding of 5-HT-1a in rat cortex also were also well fit by bilinear CLOGP with a minor dependence on a steric factor.

Upon entering their own data into C-QSAR, the Kuipers group could have obtained the following equation:

log 1/KI = 1.169(0.343)CLOGP - 1.175(0.783)BilinCLOGP + 4.943(0.683)
n = 20; r = 0.936; Q2 = 0.845; SS1 = 12.742; Dev(=) = 11
DF = 16; r2 = 0.877; S = 0.314; SS2 = 1.573; Dev(-) = 9

The program shows the results in graphical form as:

[flesinoxan analogs]

The Kuipers group concluded that the binding of compounds in the first set (those on the upslope) was directly related to hydrophobicity as mesured by their HPLC method, but binding of the second set was independent of hydrophobicity. Of course the same conclusion would have been evident without need for HPLC measurements, using CLOGP within C-QSAR. And, more importantly, they would have had a direct comparison with results of other research in this field. Actually, we noted their work soon after its publication, and the above equation (included in C-QSAR) now can be accessed to validate further research.

In demonstrating the effectiveness of their 'Stand Alone' QSAR program3 in a retrospective study of the Kuipers data, and trusting neither the hydrophobicity measurements nor the tentative conclusions of the Kuipers group, the following equation was offered as an outstanding example of the method used by the vendor's QSAR software:

Log KI = 5.19 - 0.72 X1 + 0.49 log P + 2.46 ABSQ(o,n)
(note KI not as reciprocal; thus positive coefficients lead to lower binding)

The conclusion reached is that binding is inversely related to log P (also inversely to quantum chemical calculated charges on oxygen and nitrogen). Using three descriptors for 20 analogs, a correlation with r2 = 0.86 resulted. Obviously this equation is only supported by statistics. When the data is examined correctly and validated in a comparison with other research, an entirely different conclusion is reached. That is, with these structural analogs, it is very possible that the optimal binding has been reached at log P 3.0, and synthesis of further analogs of this type can only bring modest improvement (approx. 13% in log terms).

1JACS, 1964, 86, 5175; 2Chemical Reviews, 2000, 100, 1, 1-6; 3Vendor name available on request; 4W. Kuipers, et al., J. Med. Chem., 1977, 40, 300-312

More CQSAR information (PDF; requires Adobe Acrobat reader; for more info, click here)

  • Chem-Bio Informatics and Comparative QSAR
  • C-QSAR: A General Approach to the Organization of Quantitative Structure-Activity Relationships in Chemistry and Biology
  • CQSAR Installation Instructions
  • Literature References:
    • C. Hansch and H. Gao, Chemical Reviews, 1997, 97, 2995-3059, Comparative QSAR: Radical Reactions of Benzene Derivatives in Chemistry and Biology.
    • C. Hansch, H. Gao and D. Hoekman, in "Comparative QSAR", Taylor and Francis, Wahington, D.C., J. Devillers, Ed., pp. 285-368, 1998. "A Generalized Approach to Comparative QSAR."
    • H. Gao, J. A. Katzenellenbogen, R. Garg and C. Hansch, Chemical Reviews, 1999, 99, 723-744. Comparative QSAR Analysis of Estrogen Receptor Ligands.
    • C. D. Selassie, A. J. Shusterman, S. Kapur, R. P. Verma, L. Zhang and C. Hansch, J. Chem. Soc. Perkin Trans. 2, 1999, 2729-2733. On the Toxicity of Phenols to Fast Growing Cells: A QSAR Model for a Radical-based Toxicity.
    • R. Garg, S. . Gupta, H. Gao, M. S. Babu, A. K. Debnath and C. Hansch, Chemical Reviews, 1999, 99, 3525-3601. Comparative Quantitative Structure-Activity Relationship Studies on anti-HIV drugs.
    • A. Leo and C. Hansch, Perspectives in Drug Discovery and Design, 1999, 17, 1-25. Role of Hydrophobic Effects in Mechanistic QSAR.
    • A. Leo and D. Hoekman, Perspectives in Drug discovery and Design, 1999, in press. Calculating Log P(octanol) with No Missing Fragments: The Problem of Estimating New Interaction Paramters.