Fenn, A.; Hueckelhoven, R.; Kamal, N.

Dual-organism RNA sequencing (RNA-seq) experiments, in which the transcriptomes of a host and a microbe are sequenced simultaneously, are increasingly used to study plant - microbe interactions. A central analytical goal is identifying effector proteins and their host targets through gene co-expression. Weighted Gene Co expression Network Analysis (WGCNA) is the dominant tool for gene co-expression analyses, yet its ability to recover interaction-interface genes from a merged dual-organism matrix has not been systematically characterised. Here we present a simulation framework using real gene models from Hordeum vulgare (barley) and Blumeria graminis f. sp. Hordei M.Liu & Hambl (powdery mildew) to evaluate single-network WGCNA across a gradient of plant-to-fungal library size ratios (1:1 - 20:1), three levels of co expression signal strength, and three WGCNA network construction types (signed, unsigned, signed hybrid). We embed 20 model effector genes (bridge genes) driven by a mixed host - pathogen eigengene and evaluate recovery using four metrics aligned with the biological objective: cross-species hub rank, top-decile hub enrichment, bridge gene detection rate, and bridge co-separation (the fraction of effector - target pairs co assigned to the same detected module). Across 225 simulation runs (15 conditions x 5 replicates x 3 network types), bridge genes are robustly identifiable as cross-species connectivity hubs (mean rank 0.92 versus 0.50 for module genes) but co-assignment of effector target pairs to the same module fails in 41% of runs due to scale-free topology collapse. Signal strength (n2 = 0.12) and library ratio (n2 = 0.22) are the primary determinants of co separation, while network type choice accounts for less than 2%. A read-depth bias systematically inflates pathogen gene hub ranks relative to host genes at high ratios. These results establish that the method can identify effector candidates as cross-species hubs under a broad range of conditions, but reliable co-assignment requires adequate pathogen read depth and strong co-expression signal -properties that experimental design, not analytical parameterisation, must provide.