Many different methods and techniques have been investigated for the processing and analysis of microarray gene expression profiling datasets. It is noted that the accuracy and reliability of the results are often dependent on the measurement approaches applied, and no single measurement so far is guaranteed to generate a satisfactory result. In this paper, an algorithmic fusion approach is presented for extracting genes that are predictive to clinical outcomes (survival-fatal) of diffuse large B-cell lymphoma on a set of microarray data for gene expression profiling. The approach integrates a set of measurements from different aspects in terms of the discrepancy indications and merit expectations of the gene expression patterns with respect to the clinical outcomes. A combination of statistical and nonstatistical criteria, continuous and discrete parameterizations, as well as model-based and modeless evaluations is applied in the approach. By integrating these measurements, a set of genes that are indicative to the clinical outcomes are better captured from the gene expression profiling dataset.