Huang Shuguang, Wang Yun, Chen Peining, Qian Hui-Rong, Yeo Adeline, Bemis Kerry
Statistics and Information Science, Eli Lilly & Company, Indianapolis, IN 46285, USA.
Genomics. 2004 Oct;84(4):767-77. doi: 10.1016/j.ygeno.2004.06.013.
Affymetrix's high-density oligonucleotide arrays offer an exciting technology in biomedical research. With more and more statistical involvement in every step of the process, there has been a constant effort to make sure that the expression data are appropriately extracted in the first place. According to Affymetrix GeneChip technology, each gene is represented by 11-20 oligo probe pairs; the challenge is how to extract one meaningful number, expression, from the 11-20 pairs of numbers. More specifically, there is first a need to differentiate the components of specific binding, nonspecific binding, and optical background noise in both PM and MM probes, and then an expression measure that is proportional to the true abundance of transcripts is to be derived. A new method, SUM, which sums up PM and MM values and then follows a process similar to that of RMA, is considered. The performance of SUM is investigated and compared to the three most popular methods, MAS5, dChip, and RMA. The assessments are based on a well-controlled experiment dataset that is publicly available. The results show that in several respects the performance of SUM is comparable to that of RMA and dChip, and all three of these methods show some advantages over MAS5. There is some evidence showing that SUM has higher differential sensitivity than other methods in certain situations.
Affymetrix公司的高密度寡核苷酸阵列在生物医学研究中提供了一项令人兴奋的技术。随着统计方法越来越多地参与到该过程的每一步，人们一直在不断努力确保首先能适当地提取表达数据。根据Affymetrix基因芯片技术，每个基因由11 - 20对寡核苷酸探针表示；挑战在于如何从这11 - 20对数中提取出一个有意义的数值，即表达量。更具体地说，首先需要区分PM和MM探针中特异性结合、非特异性结合以及光学背景噪声的成分，然后得出一个与转录本真实丰度成比例的表达量测量值。考虑了一种新方法SUM，它将PM和MM值相加，然后遵循与RMA类似的过程。研究了SUM的性能，并与三种最流行的方法MAS5、dChip和RMA进行了比较。评估基于一个公开可用的经过良好控制的实验数据集。结果表明，在几个方面SUM的性能与RMA和dChip相当，并且这三种方法都比MAS5具有一些优势。有证据表明，在某些情况下SUM比其他方法具有更高的差异敏感性。
