Machine Learning for Human Cancer Research

Introduction Cancer is a major public health problem. Currently, one in four deaths in the United States is due to cancer [25]. A total of 1,372,910 new cancer cases and 570,280 deaths are expected in the United States in 2005. When deaths are aggregated by age, cancer has surpassed heart disease as the leading cause of death for persons younger than 85 since 1999 [25]. The three most common cancer sites for men are lung and bronchus, colon and rectum, and prostate. The most common sites for women are breast and colorectal [25]. According to the " Centraal Bureau voor de Statistiek " (CBS, www.cbs.nl) cancer will be the leading cause of death in the Netherlands by the year 2010. These numbers justify a strong interest for this subject from researchers. There are a few widely accepted causes of cancer, for instance tobacco [12]. Although many forms of cancer are sporadic, there are some cancers with a hereditary component, such as breast cancer, which has been related to genes named BRCA1 and BRCA2 (BRCA stands for breast cancer) [19]. Cancer results from molecular events that change the properties of human cells. In cancer cells the normal control systems that prevent cell overgrowth and the invasion of other tissues are disabled. The abnormalities in cancer cells usually result from aberrations in genes that regulate cell division. Over time more genes become aberrated. This is often because the genes that make the proteins that normally repair DNA damage are themselves not functioning normally, because they are also aberrated. Consequently, aberrations begin to increase in the cell, causing further abnormalities in that cell. Some of these aberrated cells die, but other alterations may give the abnormal cell a selective advantage that allows it to multiply much more rapidly than normal cells.