Jintao Long


In this work, statistical learning approaches are exploited to discover biomarkers for Alzheimer's disease (AD). The contributions has been made in the fields of both biomarker and software driven studies. Surprising discoveries were made in the field of blood-based biomarker search. With the inclusion of existing biological knowledge and a proposed novel feature selection method, several blood-based protein models were discovered to have promising ability to separate AD patients from healthy individuals. A new statistical pattern was discovered which can be potential new guideline for diagnosis methodology. In the field of brain-based biomarker, the positive contribution of covariates such as age, gender and APOE genotype to a AD classifier was verified, as well as the discovery of panel of highly informative biomarkers comprising 26 RNA transcripts. The classifier trained by the panetl of genes shows excellent capacity in discriminating patients from control. Apart from biomarker driven studies, the development of statistical packages or application were also involved. R package metaUnion was designed and developed to provide advanced meta-analytic approach applicable for microarray data. This package overcomes the defects appearing in previous meta-analytic packages { 1) the neglection of missing data, 2) the in exibility of feature dimension 3) the lack of functions to support post-analysis summary. R package metaUnion has been applied in a published study as part of the integrated genomic approaches and resulted in significant findings. To provide benchmark references about significance of features for dementia researchers, a web-based platform AlzExpress was built to provide researchers with granular level of differential expression test and meta-analysis results. A combination of fashionable big data technologies and robust data mining algorithms make AlzExpress flexible, scalable and comprehensive platform of valuable bioinformatics in dementia research.

Document Type


Publication Date