Collagen-emulating peptides possess a high potential for applications ranging from in vivo treatment and imaging to materials design. While this potential is highest for peptides intended for heterotrimer formation, it is these peptides that are still the most difficult to control and design predictably. This difficulty is due to a dearth of known assembly-directing interactions, a wealth of competing assemblies, and an absence of reliable prediction algorithms to assess the possible arrangements. Herein we approach this problem through a cyclic method of studying amino acid substitutions and pairwise interactions in collagen-like triple helices, translating this information to stabilizing effects for use in a thermal stability prediction algorithm, using old peptides from the literature to test the algorithm’s performance, designing new peptides with the algorithm, and recycling this information to expand our understanding of sequence effects. Chapter 1 will describe the state of the field, and discuss, in-depth, eight factors which effect the stability of collagen triple helices including triple helix structure, peptide length, peptide termini, prolyl amino acids, amino acid substitutions, amino acid pairwise interactions, alternative registrations, and helical tip effects. The chapter will also examine current prediction algorithms for assessing collagen stability and if and how they assess each of the eight factors. Directions for potential improvements in the field are discussed for each of the areas. Chapter 2 details our initial investigations of amino acid substitutions and pairwise interactions in regards to collagen. We investigate the quantitative effect of known charge-pair interactions in addition to searching for new interactions by evaluating cation-π amino acid pairs and hydrophobic amino acid pairs. This information is then applied in the design of an algorithm which predicts the thermal stability of collagen mimetic peptide triple helices. Next, this algorithm is used to analyze a library of published triple helices to test its accuracy and is then retrained by machine learning to increase its accuracy and precision of prediction. Finally, the algorithm was utilized to aid in the design of a collagen-emulating heterotrimeric triple helix. This triple helix is fully characterized using CD and multiple NMR experiments to confirm that the composition and register of the folded triple helix match the intended triple helical arrangement. Chapter 3 contains an investigation of the first charge-free interaction employed for collagen triple helix design by using an amide-π pair of amino acids. This interaction was examined, along with a few others, by CD analysis of homotrimers and molecular dynamics simulations. Glutamine was determined to be the appropriate length for interacting with a phenylalanine while asparagine is too short. This interaction was applied in the design of a heterotrimer which was also fully characterized by CD and NMR. Subsequently, the data from these complete characterizations of the homotrimers and the designed heterotrimer was used to retrain the algorithm presented in chapter 2. Chapter 4 discusses a genetic algorithm created to computationally design heterotrimeric triple helices. The synthesis and characterization of five triple helices generated by the algorithm over a span of time are discussed. This work is ongoing and NMR results for two of the generated helices are not yet collected. After the conclusion of this work, the results of the five generated triple helices will be used to further train the algorithm presented in chapter 2. Chapter 5 revisits the themes of chapter 1 and describes the advances in each of the eight factors effecting the stability of collagen triple helices achieved within this thesis.