Microarray Preprocessing and mRNA/miRNA Expression Analysis

Scope and contribution. I used and adapted R/Perl analysis routines to organize supplied microarray matrices, saved annotations and sample groups. This component shows traceable probe mapping, averaging and layer-specific preparation, rather than claiming authorship of every underlying routine or a single automated integration pipeline. Image extraction and initial array processing were performed by the source-data provider, not by this workflow.

Example context and date. The source records describe rat hippocampal mRNA and miRNA measurements at 24 hours after injury. The illustrated slice retains two Sham and three TBI samples from each layer and excludes three Naive samples. Aliases S01–S08 are local to each assay; sharing an alias or group label does not establish that the two assays measured the same biological specimen. Outputs dated May 2022 establish the project year, not the date of this webpage.

Probe annotation and averaging. The mRNA source has 30,367 probes in the eight available sample columns. The saved platform annotation and eligibility rule retain 23,136 mapped probes, producing 19,273 gene-symbol rows by within-sample arithmetic averaging. All eight-column averages reconcile with the saved mapped export. Only six selected probes and six selected features are included in the interactive view; P01–P06 are presentation aliases, with full probe identifiers in details. Every displayed consolidation uses all contributing probe records for that feature.

miRNA identity and preparation. The illustrated miRNA branch has 741 probe-description rows, collapsing to 292 legacy named features. These are the saved descriptions, not silently refreshed mature-miRNA arm identifiers. After averaging repeated names and selecting the five comparison columns, the executed code applies quantile normalization to the complete 292-feature matrix. The page's replay was reconstructed using that complete matrix, with tied ranks handled consistently with the saved output, and then reduced to the selected examples. It is not normalization of the six displayed rows alone.

Units and display. The supplied values are array expression units, not sequencing counts, CPM or TPM. No executed log2 transformation is established for these branches. The mapped mRNA expression view retains its supplied scale; miRNA uses the verified quantile-normalized averaged values. Heatmap color uses log(1 + value) solely for legibility. Heatmaps share a fixed display-color scale within a layer, while the two assay layers and individual profile panels retain independent quantitative scales. The two sample-profile groups use the same ordering but not an inferred molecular pairing.

Separate statistical branch. A separate saved mRNA limma fit used the five-column probe-level matrix, not the averaged gene-symbol matrix displayed here. Its archived coefficients match differences of arithmetic group means on the supplied scale. The miRNA script also leaves its logarithmic-transform line commented out. Those coefficients cannot therefore be relabelled log2 fold changes solely because an output column is named logFC. This component does not reuse their thresholds or FDR values to assert validated differential-expression conclusions, and no statistical model was refitted.

Interpretation and interaction. Switching columns, features or preparation views inspects saved inputs and independently reconciled preprocessing values. Sample points are descriptive measurements, not evidence of causation, pathway activation or a miRNA–mRNA regulatory interaction. No authentic cross-layer target-matching execution was established in these two branches, so no regulatory edges are drawn. Unrelated tutorial, human-target, cross-study batch-correction and other result branches are excluded.