Combining Information to Develop a Safety Biomarker

• Activity Number: 37
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Sock-Cheng Lewin-Koh*+ and Nicholas Lewin-Koh and Jacqueline Tarrant
• Companies: Genentech and Genentech and Genentech
• Keywords: safety biomarker ; random forest ; factor analysis

Abstract

Significant and poorly monitorable toxicity is a common reason for drug candidate failure in preclinical drug development. In some situations, a precursor event to the toxicity may be identified and some means of monitoring the precursor event would enable better management of the toxicity. We recently encountered such a challenge with one of our molecules, but no single marker was able to monitor the liver injury satisfactorily. We proposed to combine evidence from existing information, in the hope that such a combined biomarker score may be predictive and useful for monitoring purposes. Using random forest selection and multivariate methods, we constructed a biomarker score based on selected serum markers. We assessed the effectiveness of the biomarker score and compared it against standard liver toxicity markers and another biomarker score that had been published in the literature for a similar liver injury. Results from these comparisons will be presented in this talk.