Dr. Man’s laboratory is interested in using a combined computational and molecular approach to understand pediatric cancers. One of our focuses is to construct molecular classifiers (prediction models) to predict various prognostically important subtypes of pediatric cancers based on their expression profiles. Recently, we also used a proteomic technology called Surface Enhanced Laser/ Desorption Ionization (SELDI) to profile plasma specimens from osteosarcoma patients. We are also interested in identifying important cancer signatures and biomarkers for further functional and mechanistical studies. Our current projects in the laboratory include:
Molecular classification of osteosarcoma using expression and proteomic profiling
We use high-throughput technologies, such as cDNA microarray and SELDI, to generate RNA and proteomic expression data from the clinical osteosarcoma specimens. These data are analyzed by various supervised classification algorithms to construct prediction models, which can predict clinically important phenotypes of osteosarcoma, such as drug resistance, metastatic potential, and prognosis.
Transcription network and function analysis of cancer signatures
Transcription factors are important for caner initiation, progression, and metastasis. However, the regulation of these transcription factors and how they interact with each other to form a transcriptional network is still poorly understood. One of our goals is to identify these transcription factors and their targets in osteosarcoma. Statistically significant genes in the molecular signatures identified by microarray studies are validated by other quantitative assays, such as real time PCR. These genes will be subject to further functional analysis, such as siRNA and over-expression, to understand their biological functions and underlying mechanisms.
We have generated various genomic and proteomic data on different types of specimens (tissues and plasma) in osteosarcoma. Integration of these data from different technologies will increase the possibility of identifying biologically important genes and improving the prediction accuracy of our classifiers. Microarray and proteomic datasets always contain high dimensionality while the number specimens are small. This imposes a great challenge to the bioinformatic analysis of these data to perform molecular classifications. We are interested in developing a new generation of algorithms and computational tools that can address the small sample size issue while increasing the accuracy and robustness of the classification.