Christopher Tong

Random Forest: A Classification and Regression Tool for Compound Classification and QSAR Modeling

A new classification and regression tool, Random Forest, is introduced and investigated for predicting a compounds quantitative or categorical biological activity based on a quantitative description of the compounds molecular structure. Random Forest is an ensemble of unpruned classification or regression trees created by using bootstrap samples of the training data and random feature selection in tree induction. Prediction is made by aggregating (majority vote or averaging) the predictions of the ensemble. We built predictive models for six cheminformatics data sets. Our analysis demonstrates that Random Forest is a powerful tool capable of delivering performance that is among the most accurate methods to date. We also present three additional features of Random Forest: built-in performance assessment, a measure of relative importance of descriptors, and a measure of compound similarity that is weighted by the relative importance of descriptors. It is the combination of relatively high prediction accuracy and its collection of desired features that makes Random Forest uniquely suited for modeling in cheminformatics.

Atherosclerosis Inflammation Imaging with 18F-FDG PET: Carotid, Iliac, and Femoral Uptake Reproducibility, Quantification Methods, and Recommendations

Atherosclerosis imaging with 18F-FDG PET is useful for tracking inflammation within plaque and monitoring the response to drug therapy. Short-term reproducibility of this technique in peripheral artery disease has not been assessed, and the optimal method of 18F-FDG quantification is still debated. We imaged 20 patients with vascular disease using 18F-FDG PET twice, 14 d apart, and used these data to assess reproducibility measures and compare 2 methods of 18F-FDG uptake measurement. We also reviewed the literature on quantification methods to determine the optimal measures of arterial 18F-FDG uptake for future studies. Methods: Twenty patients with vascular disease underwent PET/CT of the iliac, femoral, and carotid arteries 90 min after 18F-FDG administration. In 19 patients, repeat testing was performed at 2 wk. Coregistration and attenuation correction were performed with CT. Vessel 18F-FDG uptake was measured as both the mean and maximum blood-normalized standardized uptake value (SUV), known as the target-to-background ratio (TBR). We assessed interscan, interobserver, and intraobserver agreement. Results: Nineteen patients completed both imaging sessions. The carotid and peripheral arteries all have excellent short-term reproducibility of the 18F-FDG signal, with intraclass correlation coefficients all greater than 0.8 for all measures of reproducibility. Both mean and maximum TBR measurements for quantifying 18F-FDG uptake are equally reproducible. 18F-FDG uptake was significantly higher in the carotid arteries than in both iliac and femoral vessels (P < 0.001 for both). Conclusion: We found that both mean and maximum TBR in the carotid, iliac, and femoral arteries were highly reproducible. We suggest the mean TBR be used for tracking systemic arterial therapies, whereas the maximum TBR is optimal for detecting and monitoring local, plaque-based therapy.

Application of Breiman’s Random Forest to Modeling Structure-Activity Relationships of Pharmaceutical Molecules

Leo Breiman’s Random Forest ensemble learning procedure is applied to the problem of Quantitative Structure-Activity Relationship (QSAR) modeling for pharmaceutical molecules. This entails using a quantitative description of a compound’s molecular structure to predict that compound’s biological activity as measured in an in vitro assay. Without any parameter tuning, the performance of Random Forest with default settings on six publicly available data sets is already as good or better than that of three other prominent QSAR methods: Decision Tree, Partial Least Squares, and Support Vector Machine. In addition to reliable prediction accuracy, Random Forest provides variable importance measures which can be used in a variable reduction wrapper algorithm. Comparisons of various such wrappers and between Random Forest and Bagging are presented.

Boosting: an ensemble learning tool for compound classification and QSAR modeling.

A classification and regression tool, J. H. Friedmans Stochastic Gradient Boosting (SGB), is applied to predicting a compounds quantitative or categorical biological activity based on a quantitative description of the compounds molecular structure. Stochastic Gradient Boosting is a procedure for building a sequence of models, for instance regression trees (as in this paper), whose outputs are combined to form a predicted quantity, either an estimate of the biological activity, or a class label to which a molecule belongs. In particular, the SGB procedure builds a model in a stage-wise manner by fitting each tree to the gradient of a loss function: e.g., squared error for regression and binomial log-likelihood for classification. The values of the gradient are computed for each sample in the training set, but only a random sample of these gradients is used at each stage. (Friedman showed that the well-known boosting algorithm, AdaBoost of Freund and Schapire, could be considered as a particular case of SGB.) The SGB method is used to analyze 10 cheminformatics data sets, most of which are publicly available. The results show that SGBs performance is comparable to that of Random Forest, another ensemble learning method, and are generally competitive with or superior to those of other QSAR methods. The use of SGBs variable importance with partial dependence plots for model interpretation is also illustrated.

A novel 3-dimensional micro-ultrasound approach to automated measurement of carotid arterial plaque volume as a biomarker for experimental atherosclerosis

Improved methods for non-invasive in vivo assessment are needed to guide development of animal models of atherosclerosis and to evaluate target engagement and in vivo efficacy of new drugs. Using novel 3D-micro-ultrasound technology, we developed and validated a novel protocol for 3D acquisition and analysis of imaging to follow lesion progression in atherosclerotic mice. The carotid arteries of ApoE receptor knockout mice and normal control mice were imaged within the proximal 2mm from the aortic branch point. Plaque volume along that length was quantified using a semi-automated 3D segmentation algorithm. Volumes derived by this method were compared to those calculated using 3-D histology post-mortem. Bland-Altman comparison revealed close correlation between these two measures of plaque volume. Furthermore, using a segmentation technique that captures early positive and 33 week negative remodeling, we found evidence that plaque volume increases linearly over time. Each animal and each plaque served as its own control, allowing accurate comparison. The high fidelity anatomical registration of this protocol provides increased spatial resolution and therefore greater sensitivity for measurement of plaque wall size, an advance over 2-dimensional measures of intimal-medial-thickening. Further, 3-dimensional analysis ensures a point of registration that captures functional markers in addition to the standard structural markers that characterize experimental atherosclerosis. In conclusion, this novel imaging protocol provides a non-invasive, accurate surrogate marker for experimental atherosclerosis over the life of the entire lesion.

In vivo MRI quantification of individual muscle and organ volumes for assessment of anabolic steroid growth effects

This study aimed to develop a quantitative and in vivo magnetic resonance imaging (MRI) approach to investigate the muscle growth effects of anabolic steroids. A protocol of MRI acquisition on a standard clinical 1.5 T scanner and quantitative image analysis was established and employed to measure the individual muscle and organ volumes in the intact and castrated guinea pigs undergoing a 16-week treatment protocol by two well-documented anabolic steroids, testosterone and nandrolone, via implanted silastic capsules. High correlations between the in vivo MRI and postmortem dissection measurements were observed for shoulder muscle complex (R=0.86), masseter (R=0.79), temporalis (R=0.95), neck muscle complex (R=0.58), prostate gland and seminal vesicles (R=0.98), and testis (R=0.96). Furthermore, the longitudinal MRI measurements yielded adequate sensitivity to detect the restoration of growth to or towards normal in castrated guinea pigs by replacing circulating steroid levels to physiological or slightly higher levels, as expected. These results demonstrated that quantitative MRI using a standard clinical scanner provides accurate and sensitive measurement of individual muscles and organs, and this in vivo MRI protocol in conjunction with the castrated guinea pig model constitutes an effective platform to investigate the longitudinal and cross-sectional growth effects of other potential anabolic steroids. The quantitative MRI protocol developed can also be readily adapted for human studies on most clinical MRI scanner to investigate the anabolic steroid growth effects, or monitor the changes in individual muscle and organ volume and geometry following injury, strength training, neuromuscular disorders, and pharmacological or surgical interventions.

High-Throughput Doppler Toolbox for Preclinical Drug Development

Quantification of hemodynamics may be invaluable in a drug development setting. However, one of the challenges in the application of imaging technologies for compound screening purposes is the large volume of data in a short amount of time. This article focuses on methods developed for large-scale hemodynamic quantification, as measured from high-frequency Doppler ultrasound in rodents. An integrative semiautomated method for processing Doppler ultrasound images is described and validated. In the context of experimental biology, this toolbox allows for a comprehensive hemodynamic evaluation of in vivo physiology as the result of medical intervention, thus enabling rapid compound screening in preclinical drug development.

In vivo MRI evaluation of anabolic steroid precursor growth effects in a guinea pig model.

Anabolic steroids are widely used to increase skeletal muscle (SM) mass and improve physical performance. Some dietary supplements also include potent steroid precursors or active steroid analogs such as nandrolone. Our previous study reported the anabolic steroid effects on SM in a castrated guinea pig model with SM measured using a highly quantitative magnetic resonance imaging (MRI) protocol. The aim of the current study was to apply this animal model and in vivo MRI protocol to evaluate the growth effects of four widely used over-the-counter testosterone and nandrolone precursors: 4-androstene-3 17-dione (androstenedione), 4-androstene-3beta 17beta-diol (4-androsdiol), 19-nor-4-androstene-3beta-17beta-diol (bolandiol) and 19-nor-4-androstene-3 17-dione (19-norandrostenedione). The results showed that providing precursor to castrated male guinea pigs led to plasma steroid levels sufficient to maintain normal SM growth. The anabolic growth effects of these specific precursors on individual and total muscle volumes, sexual organs, and total adipose tissue over a 10-week treatment period, in comparison with those in the respective positive control testosterone and nandrolone groups, were documented quantitatively by MRI.

Generating hypotheses about molecular structure–activity relationships (SARs) by solving an optimization problem

An alternator pulley includes a driving member driven and rotated via a belt from an output shaft of an engine. A driving member is disposed on an inner surface of the driving member and a one-way clutch is interposed between the driving and driven member. The one-way clutch includes rollers capable of rolling in a locked side direction along which a rotating power of the driving member is transmitted to the driven member or a free side direction along which the rotating powder is interrupted. Depending on a relative speed difference between the driving member and the driven member, the rollers are biased for pressing in the locked side direction and a torque value of the pressing is set preferably to less than 4 Nm.