Clinical manifestations of rheumatoid arthritis (RA) are diversified, and based on the manifestations, the patients with RA could be classified into different patterns under traditional Chinese medicine. These patterns decide the selection of herbal prescription, and thus they can help find a subset of rheumatoid arthritis patients for a type of therapy. In the present study, we combine genome-wide expression analysis with methods of systems biology to identify the functional gene networks for the sets of clinical symptoms that comprise the major information for pattern classification. Clinical manifestations in rheumatoid arthritis were clustered with factor analysis, and two factors (similar to cold and hot patterns in traditional Chinese medicine) were found. Microarray technology was used to reveal gene expression profiles in CD4(+) T cells from 21 rheumatoid arthritis patients. Protein-protein interaction information for these genes from databases and literature data was searched. The highly-connected regions were detected to infer significant complexes or pathways in this protein-protein interaction network. The significant pathways and function were extracted from these subnetworks using the Biological Network Gene Ontology tool. The genes significantly related to hot and cold patterns were identified by correlations analysis. MAPK signalling pathway, Wnt signaling pathway, and insulin signaling pathway were found to be related to hot pattern. Purine metabolism was related to both hot and cold patterns. Alanine, aspartate, and tyrosine metabolism were related to cold pattern, and histindine metabolism and lysine degradation were related to hot pattern. The results suggest that cold and hot patterns in traditional Chinese medicine were related to different pathways, and the network analysis might be used for identifying the pattern classification in other diseases.