Collagen proteins are spread in almost every vertebrate's tissue with mechanical function. The defining feature of this fundamental family of proteins is its well-known collagen triple-helical domain. This helical domain can have different geometries, varying in helical elongation and interstrands contact, as a function of the amino acidic composition. The helical geometrical features play an important role in the interaction of the collagen protein with cell receptors, but for the vast majority of collagen compositions, these geometrical features are unknown. Quantum mechanical (QM) simulations based on the density functional theory (DFT) provide a robust approach to characterize the scenario on the collagen composition-structure relationships. In this work, we analyze the role of the adopted computational method in predicting the collagen structure for two purposes. First, we look for a cost-effective computational approach to apply to a large-scale composition-structure analysis. Second, we attempt to assess the robustness of the predictions by varying the QM methods. Therefore, we have run geometry optimization on periodic models of the collagen protein using a variety of approaches based on the most commonly used DFT functionals (PBE, HSE06, and B3LYP) with and without dispersion correction (D3ABC). We have coupled these methods with several different basis sets, looking for the highest accuracy/cost ratio. Furthermore, we have studied the performance of the composite HF-3c method and the semiempirical GFN1-xTB method. Our results identify a computational recipe that is potentially capable of predicting collagen structural features in line with DFT simulations, with orders of magnitude reduced computational cost, encouraging further investigations on the topic.