藉由DNA微陣列晶片科技比對兩組或多組以上的生物晶片來找尋差異表現基因，目前已經被廣泛的應用在生物研究中。在分析低訊噪比晶片時，分析能力的不足使得許多相同或相似的研究得到不同或是完全矛盾的結果。雖然在分析方法的改善持續的在進行，但是倚賴t檢定來尋找差異表現基因仍舊是沒有改變的。本篇文章中，我們研究了175組資料集，認為t檢定空假設應用的不正確造成了這個缺失。我們提出了一個更適當的空假設作為解決方法並發展了一個新的合適的分析方法。這個新的分析方法取名為WABE，以z檢定為基礎，它提供可靠的基因排序以及高解析力，並且不用倚賴不具統計意義的門檻。我們使用WABE重新分析一組過敏接觸性皮炎的資料，發現經由鎳刺激過後的控制組，基因產生反應的程度比病人還要強烈。這個結果與原始的研究結果不同，我們推論這個疾病是由於一個固有系統的故障而導致的。在175組資料集中，我們嚴謹的估算過去被忽略的訊息，建議其中63%-70%的現有資料需要重新檢查。The use of high-throughput expression profiling, based on technologies such as DNA microarray, to detect differentially expressed genes between two or more groups has found many applications. The analysis deficiency of data with low signal-to-noise ratios has led to results on studies of the same or similar samples that are discordant or altogether contradictory. Although efforts are continuously made to fine-tune the methodology, the reliance of such analyses on t-test for significance testing remains unchanged. We studied 175 data sets and identified the employment of t-test’s null hypothesis as the cause of deficiency; it appears to bear little relation to the reliability of the result. We proposed a more suitable null hypothesis as solution and have developed an analysis method with according designs. The method, Weighting Arrays By Errors (WABE), is based on z-test and provides a combination of reliable gene ranking and high resolving power without relying on a non-statistical cutoff. In a reanalysis of data sets on allergic contact dermatitis using WABE, we detected stronger genomic responses to nickel exposure in controls than in patients. This result contradicted the original study and led to a new inference that identifies the disease cause as malfunction of an innate system. We used the 175 data sets to conduct a severity assessment of overlooked information and concluded that 60-70% of existing data need be reexamined.
