Peptides represent a promising molecular class for drug development. They combine several strengths of small molecules (e.g. efficient tissue diffusion, low immunogenicity and access to chemical synthesis) and key properties of biologics such as monoclonal antibodies (e.g. high affinity and specificity). Most of the peptide drugs are natural products or derivatives thereof, as for example human peptide hormones or antibacterial peptides. In recent years, a range of methodologies were established to develop peptides binding to therapeutic targets de novo. Our laboratory is specialized on the in vitro evolution of bicyclic peptides by phage display. Bicyclic peptides have two macrocyclic rings that allow for binding protein targets with high affinity and selectivity. The aim of my thesis was to improve the potency, selectivity and stability of phage-selected bicyclic peptide inhibitors of coagulation factor XII (FXII) and matrix metalloproteinase 2 (MMP-2), and to test their therapeutic potential in vivo. Specifically, I planned at improving the bicyclic peptides by substituting natural amino acids with unnatural ones. As described in the following, the substitutions to unnatural amino acids had led to substantial improvements in the bicyclic peptides, and this had enabled me the evaluation of the inhibitors in disease models. In my first and second project, I aimed at improving the potency and stability of a bicyclic peptide FXII inhibitor that was previously developed in our lab by phage display, and to test the therapeutic potential of the resulting peptide in vivo. FXII has recently been identified as a promising target for safe anticoagulation therapy. In the first project, I investigated if the insertion of a single carbon atom into the macrocyclic backbone of a bicyclic peptide FXII inhibitor can improve its binding affinity. Positions within the macrocycle susceptible to atom insertion were first identified using a scanning methodology where glycine mutants were compared to ¿-alanine mutants. Upon insertion of atoms in the backbone using ¿-amino acids or homologated cysteine analogues, two peptides showed 4.7- and 2.5-fold improved Ki values. The better one blocked FXII with a Ki of 1.5 ± 0.1 nM and was more potent than the lead peptide in inhibiting the activation of the intrinsic coagulation pathway. The strategy of ring size variation by one or several atoms should be generally applicable for the affinity maturation of in-vitro-evolved cyclic peptides. In my second project, I aimed at improving further the potency of the bicyclic peptide FXII inhibitor described before, and to additionally improve its proteolytic stability. I achieved both these goals by replacing further natural amino acids to unnatural ones. Sub-nanomolar activity for human and mouse FXII (370 and 450 pM respectively) as well as a high stability (t1/2 > 128 ± 8 h in plasma) permitted preclinical evaluation of the peptide. The inhibitor efficiently blocked intrinsic coagulation in blood plasma from human, mouse and rabbit. I further demonstrated that the peptide reduced experimental thrombosis induced by ferric chloride in mice and suppressed blood coagulation in artificial lungs in rabbits, all without increasing the risk of bleeding. This shows that the optimized bicyclic peptide is a promising candidate for thromboprotection in various medical conditions. In a third project, I aimed at improving the potency and stability of a bicyclic peptide MMP-2 inhibi