Essays on QSAR History
I began my professional career at the dawn of this new era, spawned by bursts of creativity and technological advances. I was a contributor to this paradigm in those formative years and was also a chronicler of much of the early work in one branch of what is now called QSAR. I will draw upon these memories and offer them to you in the belief that you will cast these and other contributions into a meaningful and accurate historical and archival document.
Let me begin by offering a definition of what should be included in the term QSAR. In my opinion, this term should embrace attempts to relate biological activities and properties with attributes of molecules. These attributes may be classified into three groups: 1) structure - defined as a model of the form of a molecule, 2) properties - defined as measured functions of a molecule, and 3) codes - assigned indices or descriptors that catalogue a molecule or its features outside of the definitions for structure or properties. In more detail, some comments on each category are useful.
Structure is a model of an object (molecule). It is an abstraction of the whole, created to define, encode or classify one aspect of the whole molecule. Structure is not measured, it is modeled using some coherent, consistent scheme. Examples of schemes to model structure include molecular orbital-calculated electron charge distributions, energy levels and total energies, plus topological indices and conformational predictions.
Properties are attributes that are measured as a result of the intrusion of some form of energy into a molecular system. Numerical values, usually averages over a very large number of molecules arise from this process. Examples include partition coefficients, boiling points, pKa values and so on.
Codes are numerical or descriptive labels that are assigned to a molecule or fragments to represent that entity in a relationship with some property. The codes are not defined by elements of structure or properties. They are arbitrary indices that are useful for a limited span of information. The KEYS assigned to functional groups are an example. Another example are the indicator variables used in selected situations in QSAR modeling. A recent treatment of these attributes is found in: B. Testa and L. Kier " The Concept of Molecular Structure in Structure-Activity Relationships and Drug Design" Med. Res. Rev.,  35-48 (1991).
The purpose of these preliminary remarks is to insure that all of the early work leading to models of biological activity is included in the term called QSAR and historical reviews on this subject. Many contributions were made before the term was coined. Since that time there has been some tendency to limit the meaning of QSAR to property-defined molecular attributes modeling biological activity. It is a much more inclusive term and I want to elaborate here on the early work in one branch of the general paradigm.
In the decade of the 1960's, there were formed two distinct approaches to quantifying molecular attributes with biological activity. The earlier approach, with contributions from Pullman and Streitwieser in the 1950's, related molecular activity to molecular orbital-calculated values. This was followed by the development of physical property-activity relationships in 1964, pioneered by Hansch. Each of these approaches belongs under the rubric of QSAR. At this writing I want to focus on the origins of the molecular orbital approach to QSAR, describing events of the 1960's. I have asked my colleague (of 28 years), Lowell Hall to contribute to you his recollections of the early days of another structure-based approach to QSAR, namely the use of topology to define structure.
In the structure-based studies in QSAR, one must begin by identifying the pioneering work of Bernard and Alberte Pullman in the 1950's and beyond. They utilized the simple Huckel molecular orbital (M.O.) approximation to study hydrolysis, carcinogenesis, and other biochemical reactions. Much of their early work was published in a very influential book, Quantum Biochemistry (1963). Another pioneer at that time was Andrew Streitwieser who analyzed the relationship between organic chemical reactions and Huckel M. O. calculated parameters. He also published a book, Molecular Orbital Theory for Organic Chemists, (1961), that was to influence a number of scientists to pursue the modeling route to scientific understanding. I was strongly influenced by both books.
Other pioneering efforts contributed to the ability to calculate these parameters and to model molecular activities and properties. These included the first attempt at an all-valence electron M. O. method by Del Re in 1958, the first all-valence electron method capable of predicting conformational preferences by Hoffmann in 1963, and the first all-valence electron method giving reasonable charges by Pople in 1965. Many others contributed to the methodology by creating increasingly more sophisticated M. O. methods.
The early work using these methods was largely focused on biochemical reactions and interactions. Several symposia were held during this decade. These included the Menton, France series sponsored by the Pullmans, the Jerusalem symposia begun in 1967, Gordon Conferences devoted to quantum mechanical calculations, and the Sanibel Island Conferences. The main focus of these meetings was M.O. development and applications primarily in chemical reactivity. At each symposium, there was always an invited biologist or two who would attempt to relate the theoretical methods to explain a biological phenomena. I found myself in this role on several occasions. Attempts to relate structure to drug activity was a presentation usually on the last day of the meeting and it would appear as the last chapter in the symposium volume.
At one of the Sanibel meetings in the late 1960's an organization was formed called the International Society of Quantum Biology. I was on the board of this organization for two years. I wonder if this was not the antecedent or direct stimulus for the formation of the QSAR Society?
There were a number of scientists, impressed by the work of the Pullmans and Streitwieser who ventured into the drug molecule structure-activity realm using these methods. Some interesting work was published by a handful of investigators but it was not warmly received at that time by the medicinal chemists, the quantum people or the property-activity modelers. One significant contribution occurred in 1967 when the first M. O. calculation of the preferred conformation of a drug molecule was reported. The molecule was acetylcholine, Kier, (1967). The prediction was later confirmed experimentally. Further studies in this series on other transmitters and drugs led the way to the predictions of pharmacophores using the technique labeled "receptor mapping". The prediction of preferred conformation is now an automatic procedure that all modelers accomplish by clicking on an icon to "minimize".
The first symposium devoted primarily to applications of M.O. studies on drug molecules was organized by Kier in 1969 in Seattle, Wash. This was sponsored by Battelle Memorial Institute. All of the investigators using M.O. in QSAR studies at that time were in attendance. The speakers included: Art Cammarata, Bill Purcell, Soloman Snyder, Jack Green, Brock Neely, Arnie Wohl,
The latest work in structure-based QSAR was presented at the symposium. The symposium book from this meeting, the first book dealing with drug QSAR, was published the following year: Molecular Orbital Studies in Chemical Pharmacology, L. B. Kier, ed. (1970). The QSAR studies in the 60's using molecular orbital theory were summarized in a primer on the subject, published in 1971: Molecular Orbital Theory in Drug Research, L. B. Kier (1971).
These were the events in the first decade of structure-based QSAR studies that I recall. I can look for memorabilia and more details as you wish. I hope that I have not omitted anyone but that will surface as you work on this project. Lowell and I will continue the effort to assist you and look forward to your comments.