The grand tour and projection pursuit are two methods for exploring multivariate data. We show how to combine them into a dynamic graphical tool for exploratory data analysis, called a projection pursuit guided tour. This tool assists in clustering data when clusters are oddly shaped and in finding general low-dimensional structure in high-dimensional, and in particular, sparse data. An example shows that the method, which is projection-based, can be quite powerful in situations that may cause grief for methods based on kernel smoothing. The projection pursuit guided tour is also useful for comparing and developing projection pursuit indexes and illustrating some types of asymptotic results.

A major issue in designing drugs as antagonists at the glycine site of the NMDA receptor has been to achieve good in vivo activity. A series of 4-hydroxyquinolone glycine antagonists was found to be active in the DBA/2 mouse anticonvulsant assay, but improvements in in vitro affinity were not mirrored by corresponding increases in anticonvulsant activity. Here we show that binding of the compounds to plasma protein limits their brain penetration. Relative binding to the major plasma protein, albumin, was measured in two different ways: by a radioligand binding experiment or using an HPLC assay, for a wide structural range of glycine/NMDA site ligands. These measures of plasma protein binding correlate well (r = 0.84), and the HPLC assay has been used extensively to quantify plasma protein binding. For the 4-hydroxyquinolone series, binding to plasma protein correlates (r = 0.92) with log P (octanol/pH 7.4 buffer) over a range of log P values from 0 to 5. The anticonvulsant activity increases with in vitro affinity, but the slope of a plot of pED50 versus pIC50 is low (0.40); taking plasma protein binding into account in this plot increases the slope to 0.60. This shows that binding to albumin in plasma reduces the amount of compound free to diffuse across the blood-brain barrier. Further evidence comes from three other experiments: (a) Direct measurements of brain/blood ratios for three compounds (2, 16, 26) show the ratio decreases with increasing log R. (b) Warfarin, which competes for albumin binding sites dose-dependently, decreased the ED50 of 26 for protection against seizures induced by NMDLA. (c) Direct measurements of brain penetration using an in situ brain perfusion model in rat to measure the amount of drug crossing the blood-brain barrier showed that compounds 2, 26, and 32 penetrate the brain well in the absence of plasma protein, but this is greatly reduced when the drug is delivered in plasma. In the 4-hydroxyquinolones glycine site binding affinity increases with lipophilicity of the 3-substituent up to a maximum at a log P around 3, then does not improve further. When combined with increasing protein binding, this gives a parabolic relationship between predicted in vivo activity and log P, with a maximum log P value of 2.39. Finally, the plasma protein binding studies have been extended to other series of glycine site antagonists, and its is shown that for a given log P these have similar protein binding to the 4-hydroxyquinolones, except for compounds that are not acidic. The results have implications for the design of novel glycine site antagonists, and it is suggested that it is necessary to either keep log P low or pKa high to obtain good central nervous system activity.

No abstract

Many graphical methods for displaying multivariate data consist of arrangements of multiple displays of one or two variables; scatterplot matrices and parallel coordinates plots are two such methods. In principle these methods generalize to arbitrary numbers of variables but become difficult to interpret for even moderate numbers of variables. This article demonstrates that the impact of high dimensions is much less severe when the component displays are clustered together according to some index of merit. Effectively, this clustering reduces the dimensionality and makes interpretation easier. For scatterplot matrices and parallel coordinates plots clustering of component displays is achieved by finding suitable permutations of the variables. I discuss algorithms based on cluster analysis for finding permutations, and present examples using various indices of merit.

REFERENCESLinked references are available on JSTOR for this article: https://www.jstor.org/stable/1390844?seq=1&cid=pdf-reference#references_tab_contents You may need to log in to JSTOR to access the linked references.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.The grand tour and projection pursuit are two methods for exploring multivariate data. We show how to combine them into a dynamic graphical tool for exploratory data analysis, called a projection pursuit guided tour. This tool assists in clustering data when clusters are oddly shaped and in finding general low-dimensional structure in high-dimensional, and in particular, sparse data. An example shows that the method, which is projection-based, can be quite powerful in situations that may cause grief for methods based on kernel smoothing. The projection pursuit guided tour is also useful for comparing and developing projection pursuit indexes and illustrating some types of asymptotic results.