In high-throughput cancer genomic studies, markers identified from the analysis of single data sets often suffer a lack of reproducibility because of the small sample sizes. An ideal solution is to conduct large-scale prospective studies, which are extremely expensive and time consuming. A cost-effective remedy is to pool data from multiple comparable studies and conduct integrative analysis. Integrative analysis of multiple data sets is challenging because of the high dimensionality of genomic measurements and heterogeneity among studies. In this article, we propose a sparse boosting approach for marker identification in integrative analysis of multiple heterogeneous cancer diagnosis studies with gene expression measurements. The proposed approach can effectively accommodate the heterogeneity among multiple studies and identify markers with consistent effects across studies. Simulation shows that the proposed approach has satisfactory identification results and outperforms alternatives including an intensity approach and meta-analysis. The proposed approach is used to identify markers of pancreatic cancer and liver cancer.