Compartive proteome, transcriptome and co-expression network data set of six plant species

Proteome: MS data was searched with the MaxQuant software (version 1.5.7.5). Searches were performed against the following UniProt protein sequence databases (v20150827): Medicago truncatula, Arabidopsis thaliana, Solanum tuberosum, Oryza sativa, Physcomitrella patens, and Zea mays. Searches used the default precursor mass tolerances (20 ppm first search and 4.5 ppm main search) and a product mass tolerance of 0.35 Da. The in silico digest was set to specific tryptic cleavage and a maximum of two missed cleavages. The fixed modification specified were carbamidomethylation of cysteine residues, and variable modifications were oxidation of methionine and acetylation of protein N-terminus. Peptides and proteins groups were both filtered to a 1% FDR. Label-free quantification was performed within MaxQuant using MaxLFQ. Protein groups were screened for "Reverse" and "Contaminant" identifications. The final protein levels are the LFQ intensity values. Transcriptome: The compendia of transcriptomic datasets for 6 plant species were created using publicly available RNA-seq dataset from Sequence Read Archive. We used the Curse and Prose suite (http://bioinformatics.psb.ugent.be/webtools/Curse/) to construct these expression compendia. First, we searched for datasets corresponding to "RNA-seq" and "Transcriptome analysis" for each species, and manually excluded the experiments which contained mutant, knock-out, and transgenic lines by curating the metadata in the Curse interface. We passed the chosen experiments and genomic sequence files to the Prose tool, which downloads data, performs quality control, and quantifies transcript expression to normalized TPM values. To change the transcript level expression to gene level, we summed up the multiple values of transcript expression for one gene. Co-expression network: We defined a species-specific graph as a fully-connected weighted graph, where the weight corresponds to the Gaussian kernel-based similarity between two gene expression profiles. First, we collected genes that are defined in the orthogroups of 6 species in the previous step (see method in the paper). Then we generated a distance matrix by calculating the pairwise Euclidean distance between two genes of the transcriptome compendium, and finally we transformed it to the similarity matrix using the Gaussian kernel to map the nearer distance to higher similarity.