This document describes a multi-objective simulated annealing (MOSA) based clustering method for cancer diagnosis and classification using gene expression data. The method uses MOSA to identify optimal clustering solutions based on multiple objectives like minimizing intra-cluster distances and maximizing inter-cluster distances. It is tested on three cancer gene expression datasets and able to accurately cluster samples and identify relevant gene markers for different cancer types. The method performs better than other algorithms like MOGASVM in terms of clustering accuracy and identifies clinically relevant gene markers for cancer subtypes.