Disease Subtype Discovery Research Articles

Shaping the Transformation of Pharmaceutical Science

This final issue of Pharmaceutical Research in 2008 is dedicated to the celebration of the journal’s 25th anniversary. This particular issue is a collage of (a) selected invited papers for the symposium entitled, “25th Anniversary of Pharmaceutical Research: Impact on Drug Development and Roadmap to Future Innovations” at the AAPS meeting in Atlanta and (b) contributions from the authors at large whose work collectively defines the future direction of Pharmaceutical Research. The themed sections are illustrative of the indispensable role of integrated research in modern science and of the translation of basic research to disease management. Pharmaceutical science is translational by its very nature. My goal as editor-in-chief was to position Pharmaceutical Research as the forum of choice for assuring the scientific quality as well as for archiving translational research in pharmaceutical science. In my view, the journal is a rich source of information and case studies on the integration of fundamental principles in chemistry, biology, and engineering for creating and evaluating innovative products or protocols that meet the needs of a healthy society. Transport, formulation, biopharmaceutics and pharmacokinetics, simulation and mathematical modeling, and drug delivery are the building blocks of Pharmaceutical Research. They comprise the tool kit for creating innovative solutions to challenging entities. These entities range from small organic molecules to stem cells, which are on their way to becoming life-enrichment medicine. The top 25 papers from 1984 to the present, listed in Table I, span the area defined by the keywords mentioned earlier. The range of average citations per year is from 8.95 to 48.29. All but two of these papers were published in the decade of 1990–1999; the other two papers were published in 1998 (ranked no. 15) and in 2002 (ranked no. 22). Pharmaceutical Research continues to flourish because of your support. To date, 13,985 authors have published in a total of 6,018 papers in the journal. Over 50% of the work was conducted in the laboratories of the 25 investigators listed in Table II. The top five contributors in this group are R.H Guy. G.L. Amidon, Y. Sugiyama, R. Borchardt, and V. Shah in that order. Table II also lists the top 25 institutions as well as the top 25 countries in number of papers published. The US Food and Drug Administration and three pharmaceutical companies (Upjohn, Genentech, and Pfizer) are sharing the top 25 institutions with universities in the United States, Japan, Israel. Sweden, and the Netherlands. In terms of countries, the United States contributes more than half of the published papers, followed by Japan (11.6%), and England (5.42%). The up and coming countries in Asia (South Korea, China, India, Singapore and Taiwan) together account for 4% of all published manuscripts in Pharmaceutical Research (Table III). The pharmaceutical world is dynamic, and realignment in the lineup of major contributors, institutions, and countries to the toptiered has already started. Although the United States and Japan together still contribute to 60% of all published manuscripts in recent years (2003–present) (Table II), only one of the top five most cited work is from the United States. The underlying cause and the strategy to reverse such a trend are very much needed. With 676 citations to date, the paper by Amidon et al. (paper no. 1 in Table I), which formalized the biopharmaceutic drug classification (known for short as BCS), is the most highly cited paper in the history of the journal. This citation classic presented an elegant way to select the fundamental drug properties deemed important in bioavailability. The lesson learned from such a strategy undoubtedly will exert a profound impact on the strategy of drug product development by drug manufacturers and on the approach adopted by drug regulators to examine drug applications. In so doing, the BCS concept serves as an effective interface between the business sector and the government agency charged with safeguarding the health of the public from drug misadventures. Other topics that are on the more recent top 25 list include protein formulations; drug transporters; prediction, experimental evaluation, and modification of epithelial drug permeability; and polymeric carriers for oligonucleotides and genes. We are living in a world of unprecedented change. Change knows no geographical, disciplinary or national boundaries. First, the internet has created a cost effective way for sharing information with anyone, anywhere, anytime. It is already changing the business model of publishing. Second, the confluence of biotechnology, nanotechnology, and informational technology has enabled scientists to tackle complex scientific questions. Discovery of disease subtypes, hence the need for more sophisticated treatment, may be on the rise. Third, the resolution of such questions as well as an improved understanding of disease at the genetic, cellular and molecular levels will accelerate the development of personalized

Penalized model-based clustering with cluster-specific diagonal covariance matrices and grouped variables

Clustering analysis is one of the most widely used statistical tools in many emerging areas such as microarray data analysis. For microarray and other high-dimensional data, the presence of many noise variables may mask underlying clustering structures. Hence removing noise variables via variable selection is necessary. For simultaneous variable selection and parameter estimation, existing penalized likelihood approaches in model-based clustering analysis all assume a common diagonal covariance matrix across clusters, which however may not hold in practice. To analyze high-dimensional data, particularly those with relatively low sample sizes, this article introduces a novel approach that shrinks the variances together with means, in a more general situation with cluster-specific (diagonal) covariance matrices. Furthermore, selection of grouped variables via inclusion or exclusion of a group of variables altogether is permitted by a specific form of penalty, which facilitates incorporating subject-matter knowledge, such as gene functions in clustering microarray samples for disease subtype discovery. For implementation, EM algorithms are derived for parameter estimation, in which the M-steps clearly demonstrate the effects of shrinkage and thresholding. Numerical examples, including an application to acute leukemia subtype discovery with microarray gene expression data, are provided to demonstrate the utility and advantage of the proposed method.