Traditional online toxicity analysis focuses on individual users, overlooking structural dynamics within online communities. We propose the Weighted Focal Structure Analysis (WFSA) algorithm to identify focal toxic structures (FTSs): densely interconnected node groups that intensify toxic discourse. WFSA demonstrates significant gains in detecting toxic influence structures, validated using F1 scores and standard metrics. We apply SIR, SEIR, and SEIZ models with 1,500 Monte Carlo simulations to assess toxicity propagation by FTSs versus influential toxic individuals (ITIs). Results show FTSs significantly outperform highly central individuals in propagating toxicity across network topologies. SEIZ achieves the lowest macro-error, confirming predictive robustness. Targeting FTSs provides a scalable, effective strategy to mitigate toxicity, advancing network-based approaches for healthier digital communities.