Expression levels measured in microarrays of oligonucleotide probes have now been adapted as a high throughput approach for identifying DNA sequence variation between genotypes, referred to as single feature polymorphisms (SFPs). Although there have been increasing interests in this approach, there is still need for improving the algorithm in order to achieve high sensitivity and specificity especially with complex genome and large datasets, while maintaining optimal computational performance. We obtained microarray datasets for expression profiles of two rice cultivars and adapted a median polish method to detect SFPs. The analysis identified 6,655 SFPs between two the rice varieties representing 3,131 rice unique genes. We showed that the median polish method has the advantage of avoiding fitting complex linear models thus can be used to analyze complex transcriptome datasets like the ones in this study. The method is also superior in sensitivity, accuracy and computing time requirement compared with two previously used methods. A comparison with data from a resequencing project indicated that 75.6% of the SFPs had SNP supports in the probe regions. Further comparison revealed that SNPs in sequences immediately flanking the probes also had contributions to the detection of SFPs in cases where the probes and the targets had perfectly matched sequences. It was shown that differences in minimum free energies caused by flanking SNPs, which may change the stability of RNA secondary structure, may partly explain the SFPs as detected. These SFPs may facilitate gene discovery in future studies.