An automatic molecular diagnosis approach based on boosting classification for ALS disease

Protein profiling in tissues and fluids in disease and pathological controls is used for molecular diagnosis. Proteomics will play an important role in diagnosis with therapeutics and personalized healthcare. We carried out a new robust diagnostic method based on boosting-based alternating decision tree to improve classification accuracy. It generates classification rules, which are often smaller and easier to interpret. It often gives most discriminate features, which can be utilized as biomarkers for diagnostic purpose. Also, it has a nice feature of giving a measure of prediction confidence. We carried out this strategy in ALS (Amyotrophic lateral sclerosis) disease data by different SELDI MS experimental procedures. The cross-validation and ROC analysis results indicated that good prediction capacity was obtained based on SELDI results through WCX procedure, while SAX2 and IMAC SELDI procedures did not give sufficient discriminate information for ALS disease. Our molecular diagnosis method provides an efficient way to distinguish ALS disease from neurological controls. Its results are expressed in a simple and straightforward alternating decision tree format or conditional format. We identified most discriminate peaks in proteomic data, which can be utilized as biomarkers for diagnosis. It indicated that our diagnosis method is very useful and effective in classification in neurological diseases like ALS It will have broad application in molecular diagnosis through proteomics data analysis and personalized medicine in this post-genomic era.