A Bernoulli/Left-Censored Lognormal Mixture Model for Activity of the Protease of HIV-1 as a Function of Amino Acid Characteristics

• Activity Number: 43
• Type: Invited
• Date/Time: Sunday, August 6, 2006 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Paul W. Stewart*+
• Companies: The University of North Carolina at Chapel Hill
• Keywords: HIV ; mixture model ; protease ; AIDS ; left censored

Abstract

In a recent study by the UNC CFAR, we developed a mixture model for mean level of HIV-1 protease activity as a function of structural characteristics of the protease. The protease is an enzyme required for HIV replication. Previous protease research resulted in a class of FDA-approved drugs: protease inhibitors. As mutations of the protease occur, various amino acids are substituted for the amino acids usually found in the protease. The choices of amino acids in key positions have been manipulated experimentally, yielding observed levels of protease activity that range from 0% to 100%, subject to a lower limit of detection. Some configurations of the protease rendered activity impossible. A statistical model developed for these data will be described and issues of estimation, inference, and diagnostics will be discussed.