Indicate in the comments box below whether you are happy with the changes made or if the manuscript is unacceptable.

Comments on revised manuscript

The revision is improved and should be accepted. I only have a few editing suggestions this round: 1. "Poly Pipeline: A Universal Spatial Transcriptomics Workflow Validated Across Polyploid and Diploid Organisms" would be a tighter title. 2. Several sentences still overstate the pipeline's polyploid specificity (e.g., "specifically tailored to overcome the complexity of polyploid plant genomes"; "optimization for polyploid organisms … had not yet been specifically addressed"). The KNN-based metacell aggregation underlying this claim mitigates data sparsity, a universal problem in single-cell and spatial transcriptomics, and does not itself address homoeologous read assignment, subgenome identity, or dosage normalization; in the absence of a control comparison, a polyploid-specific advantage cannot be claimed. These statements, together with the unsupported claim that the Spatial Leiden algorithm "proved superior," should be reworded or removed. 3. three miscited references [9], [16], [25] 4. My student tried and found the submitted code does not appear to actually call hdWGCNA or implement its KNN algorithm, so please check the github codes: "I located the network-analysis code at lines 4509–4692 of the main program, 2_COMPLETE_ANALYSIS.sh. This block dynamically generates bin/SCRIPT_NETWORK_ANALYSIS.r. The dynamically generated script does attach the hdWGCNA package through library(hdWGCNA). However, I could not identify any subsequent call to an hdWGCNA function, such as SetupForWGCNA, MetacellsByGroups, MetaspotsByGroups. More importantly, the step labelled “Metacells” appears to use k-means rather than KNN aggregation. The author used the k-means algorithm directly to divide the cells into 25 clusters, which createing 25 fixed k-means-derived pseudobulk samples. It is fundamentally different from the KNN-bagging procedure used by hdWGCNA. The subsequent network is constructed using conventional WGCNA functions, including pickSoftThreshold, adjacency, TOMsimilarity, cutreeDynamic, and mergeCloseModules. A further concern is that the network input is restricted to the HVGs selected during the upstream analysis. Using HVGs for co-expression network construction is not necessarily invalid and hdWGCNA itself permits variable-feature selection. However, this implementation cannot support the claim that the network module “rescues” low-expressed duplicated genes that would otherwise be removed during HVG filtering. I recommended that the author add an optional gene input parameter before running the hdWGCNA module, allowing users to choose whether the gene set passed to hdWGCNA is HVG, all genes, genes expressed in a certain proportion of cells, or a custom gene set."