Effective and reliable prediction for ecotoxicity, especially when affecting different levels of trophic chains, including humans, is increasingly gaining even more prominence as ecosystems face new threats and challenges, as that posed by the per- and poly-fluoroalkyl substances (PFAS). Toxicological prediction of PFAS in aquatic organisms, such as zebrafish, can be efficiently achieved through computational ecotoxicological approaches which are fully aligned with the state-of-the-art of new approach methodologies (NAMs) and current regulatory recommendations. Specifically in this work, the PFAS toxicodynamics interaction on the zebrafish mitochondrial voltage-dependent anion channel (zfVDAC2) was evaluated, mimicking in silico the PFAS bioaccumulation in low-concentration by integrating structure-based virtual screening (SB-VS) and predictive quantitative structure-activity(mitotoxicity) relationship (QSAR) methodologies (e.g., 2D/3D-QSAR) to address mechanistic aspects of PFAS toxicity. The best ranked PFAS pose docked in zfVDAC2 exhibits a ΔG-binding affinity higher than the ATP, i.e., the native substrate of the zfVDAC2 channel, with prevalence of van der Waal interactions, followed by fluorine (F)-halogen-bonds and finally hydrogen-bonds interactions. Mitochondrial ATP-transport blocking is thus suggested to be linked with local-flexibility perturbations in the zfVDAC2. Similarly, the obtained 2D/3D- QSAR models point out the packing density index as the most significant PFAS molecular descriptor to induce toxicity in the zfVDAC2, and mainly involving van der Waal interactions. The predictive and statistical performance of these models further indicate its NAM relevance regarding PFAS risk assessment while highlighting its interoperability and extrapolation capability for the ecotoxicological evaluation of other families of compounds.