Predicting the Regioselectivity of Mechanochemical Bond Scission in Complex Molecules

Mechanical force redistributes reactivity through anisotropic, topology‐dependent load transmission. Predicting which covalent bond in a complex molecule ruptures first under tension remains a central problem in mechanochemistry, biomaterials and chemically recyclable polymers. Bond dissociation energy is an incomplete proxy because it describes zero‐force thermodynamic stability, whereas mechanochemical site selectivity is governed more directly by the peak force sustained along a prescribed loading path. Here we introduce covalent‐bond peak force network (CBPFNet), a graph‐attention model that predicts covalent‐bond peak force (CBPForce) from relaxed molecular structures through a virtual‐stretching workflow and avoids the hundreds of constrained geometry optimisations typically required for each candidate bond. In the CHON benchmark studied here, CBPFNet reproduces bond‐ranking trends from density functional theory (DFT) and can be used for bond‐level screening. Post hoc attribution analyses indicate that rupture selectivity depends on both bond identity and local stiffness contrast, with relatively rigid segments concentrating tensile stress on softer adjacent bonds. In oligopeptides, the model recovers the DFT weakest bond and highlights a serine‐containing motif as a possible topology‐dependent weak point under the same protocol. CBPFNet therefore extends screening beyond equilibrium bond stability toward bond‐resolved peak‐force ranking in sustainable polymer design and biomaterials.