Refseq Transcript Set. Sequencing-by-synthesis: Cerebellum samples were sequenced using Illumina Genome Analyzers with modifications for mRNA samples (Lister et al., 2008; Morin et al., 2008)(Mortazavi et al., 2008): Following quality assessment using a Bioanalyzer 2100 (Agilent Inc., Santa Clara, CA; Table 1), poly A+ RNA was isolated from 5 – 10 μg total RNA by two rounds of oligo-dT selection (Invitrogen Inc., Santa Clara, CA). mRNA was annealed to high concentrations of random hexamers and reverse transcribed. Following second strand synthesis, end repair, and A-tailing, adapters complementary to sequencing primers were ligated to cDNA fragment ends. Resultant cDNA libraries were size fractionated on agarose gels, 200 bp fragments excised, and amplified by 15 cycles of polymerase chain reaction. Following quality assessment using a Bioanalyzer 2100, single-stranded cDNA-adapter fragments were randomly annealed to the surface of a flow cell in a cluster station (Illumina Inc., San Diego, CA) via primers complementary to the adapters and incubated under conditions fostering annealing of the ends of cDNA-adapter fragments to adjacent complementary primers. Primers with annealed cDNA-adapter fragments were extended with DNA polymerase and unlabelled dNTPs in a solid-phase “bridge amplification”. Resultant double-stranded products were denatured, yielding 2 single stranded fragments. The latter steps were repeated for 35 cycles, generating ~40 million clusters of clonal amplicons. Subsequently, 32 – 36 cycles of sequencing-by-synthesis chemistry were performed in Illumina Genome Analyzer instruments with 4 dNTPs featuring cleavable dyes and reversible terminators. Following each base extension, 4 images (one for each nucleotide) are taken upon laser excitation. Incorporation of the next base occurred after removal of the blocked 3’ terminus and fluorescent tag of the previously incorporated nucleotide. Read Alignment-based gene expression profiling: Reads were aligned to the human genome and RefSeq transcript databases (Pruitt et al., 2005) using the algorithm GMAP (Wu and Watanabe, 2005) and the software system Alpheus (Sugarbaker et al., 2008), with adjustments for short SBS reads (oligomer overlap interval = 3nt, identities > 34/36 or 94%). Only the highest scoring alignment(s) was retained. Reads with a single best alignment or with equally good alignments to alternative transcripts of the same gene were considered uniquely aligned. Aligned read frequencies (per million aligned reads) were calculated for each sample and gene using Alpheus (Sugarbaker et al., 2008). Statistical Analysis: Array hybridization signals and read frequencies were log10 transformed prior to evaluation of inter-sample differences. Overlayed kernel density estimates, univariate distribution results, Mahalanobis distances, correlation coefficients of pairwise sample comparisons, unsupervised principal component analysis (by Pearson product-moment correlation) and Ward hierarchical clustering of Pearson product-moment correlations of read frequencies were performed with JMP Genomics, Version 3.2 (SAS Institute, Cary, NC). Decomposition of principal components of variance of array signals and read frequencies, FDR corrected analysis of variance (with inclusion of the non-diagnosis components of variance discussed above as fixed effects) and chi squared comparisons of GO annotations were also performed with JMP Genomics, Version 3.2. Patient, sample, and experimental parameters examined to quantify sources of variability in measurements were age, alignment percentage, brain pH, cause of death, cluster station, diagnosis, library creator, number of reads, post-mortem interval, psychotropic medication, race, read length, read quality score, RNA isolation date, RNA integrity number, sample storage duration, sequencing instrument and year sequenced.
Less...
More...