2.7.3. Statistical Analysis

Five biologically replicated comparisons of contrasting cohorts (control and water deficit) were used, with target-decoy searching employed to limit aggregate PSM error to 1%. The spectral count data were compared in an R statistics script, with a minimum information criterion of 10 spectra per protein being set, after which a Quasi-Poisson regression was conducted with treatment (well-watered/water deficit) and species variables. Values for multiple testing were corrected via the Benjamini–Hochberg FDR method [33]. If a protein had a q-value < 0.05, it was considered significantly different, with 5% of the claimed changes expected to be false positives.

One-way analysis of variance (ANOVA) was used to analyze the physiological and biochemical results and their significance was determined using the Tukey–Kramer method at a 5% significance level.

#### 2.7.4. Gene Ontology and KEGG Analysis

Differentially expressed proteins were functionally annotated using the Blast2GO program implemented in the OmicsBox v2.2.4 software [34]. The sequence information of the differentially expressed orthologs were obtained from the UniProtKB website in FASTA format on 19 September 2022. Protein sequences were searched, using the basic local alignment search tool (BLAST), against sequences in NCBI (National Center for Biotechnology Information) via the BlastP search algorithm to determine similarity matches. The BLAST search was carried out using the default parameters with a maximum of 20 hits, at an expectation value of 1.0 × <sup>10</sup>−3, with 33 as the high-scoring segment pair (HSP) length cutoff and 0 as the HSP-hit coverage, with application of a low complexity filter. Sequences that received the best BLAST hits were mapped and annotated using default settings (annotation cutoff 55, Go weight 5, e-value 1.0 × <sup>10</sup>−6, HSP-hit coverage cutoff 0 and hit filter 500). Proteins were annotated according to gene ontology (GO) by 'level 2' on the basis of molecular function (MF), biological process (BP) and cellular component (CC). Protein sequences were scanned for conserved domains against signatures in InterPro using the InterProScan tool, which was an inbuilt program of Blast2GO. Annotated sequences were linked to metabolic pathways via Enzyme Commission (EC) numbers using the KEGG (Kyoto Encyclopedia of Genes and Genomes) extension of Blast2GO.

#### **3. Results**
