Transactive response DNA binding protein 43 (TDP-43), an alternative-splicing regulator, can specifically bind the TG-rich DNAs, which is associated with a range of neurodegenerative diseases. Molecular dynamics simulation, although powerful in exploring inter-molecular interactions, is time-consuming, and moreover it is difficult to sample sufficiently the conformations for the system with large conformational changes to study the allosteric behavior. Here, we utilize a coarse-grained, elastic potential-based Gaussian network model (GNM) to characterize the interacting dynamics between human TDP-43 and DNA. Furtherly, using our group’s previously proposed thermodynamic cycle method based on GNM, we identify the key residues for DNA binding whose perturbations induce a large change in their binding free energy. The results reveal that upon DNA binding, an evident loss of flexibility occurs to TDP-43’s loop1 and loop3 segments rich in positively charged resides, which indicates their induced fit role in TDP-43-DNA recognition and interactions. Additionally, the thermodynamic cycle method identifies not only the residues important for DNA specific binding, but also the ones far away from the binding interface but critical for the conformational changes of TDP-43 caused by the DNA binding. This study is helpful for the understanding of the specific interaction between TDP-43 and DNA, and can provide important information for the related drug design. In addition, this method can be easily extended to other protein-nucleic acid interacting dynamics studies.