Is there any difference in using "manual" WGCNA vs. using the BlockwiseModules function in terms of the quality of the results?

Short answer: No inherent quality difference—they use the same core WGCNA steps (soft-thresholded adjacency → TOM → hierarchical clustering → dynamic tree cut → module merging). Differences you see usually come from parameters and memory strategy, not from “manual vs. blockwise” per se.

Blockwise strategy: blockwiseModules() partitions genes into blocks to fit memory/parallelize. If your data fit in RAM and you set one big block, results ≈ a manual end-to-end run with the same parameters.

When “manual” helps: (1) you want full control at each step (custom adjacency/TOM, outlier handling, bespoke clustering or merging); (2) you’re iterating interactively to tune parameters on a subset, then locking them for the full run; and (3) you need to recalculate/merge modules with custom criteria or perform special QC between stages.

Hope this helps.