Sources of variability and effect of experimental approach on expression profiling data interpretation

摘要: We provide a systematic study of the sources variability in expression profiling data using 56 RNAs isolated from human muscle biopsies (34 Affymetrix MuscleChip arrays), and 36 murine cell culture tissue (42 U74Av2 arrays). studied 28 subjects as well myogenic cultures, muscle, spleens. Human arrays (4,601 probe sets) microarrays were used for profiling. profiled both singly, mixed groups. Variables included heterogeneity, cRNA production, patient diagnosis, GeneChip hybridizations. found that greatest source was often different regions same biopsy, reflecting variation type content even relatively homogeneous such muscle. Inter-patient also very high (SNP noise). Experimental (RNA, cDNA, cRNA, or GeneChip) minor. Pre-profile mixing samples effectively normalized intra- inter-patient variation, while retaining degree specificity individual profiles (86% statistically significant differences detected by absolute analysis; 85% 4-pairwise comparison survival method). Using unsupervised cluster analysis correlation coefficients 92 RNA on 76 oligonucleotide microarrays, we experimental error not unwanted experiments. Major use small biopsies, particularly humans where there is substantial