Comparison of High-Level Microarray Analysis Methods in the Context of Result Consistency

摘要: Motivation When we were asked for help with high-level microarray data analysis (on Affymetrix HGU-133A microarray), faced the problem of selecting an appropriate method. We wanted to select a method that would yield "the best result" (detected as many "really" differentially expressed genes (DEGs) possible, without false positives and negatives). However, life scientists could not us – they use their "favorite" special argumentation. also did find any norm or recommendation. Therefore, decided examine it our own purpose. considered whether results obtained using different methods analyses Significant Analysis Microarrays, Rank Products, Bland-Altman, Mann-Whitney test, T test Linear Models Microarray Data be in agreement. Initially, conducted comparative on eight real sets from experiments (from Array Express database). The surprising. On same array set, set DEGs by significantly different. applied artificial determined some measures allow preparation overall scoring tested future recommendation. Results We found very low level concordance sets. number common all six fixed sets, checked sets) ranged 6 433 (22,283 total readings). Results better than those data. fully satisfying. scored accuracy, recall, precision, f-measure Matthews correlation coefficient. Based scoring, SAM LIMMA. TT acceptable. worst was MW. study, recommend: 1. Carefully taking into account need study when choosing method, 2. Making more one then only are (which seems reasonable) 3. Being careful (while summarizing facts) about genes: discover DEGs.