Development of a High-Throughput Computational Framework for Peptide Discovery with Application to Blood-Brain Barrier Penetration

Abstract

The blood–brain barrier (BBB) remains a major obstacle to the delivery of therapeutics into the central nervous system. Identifying short peptides capable of crossing this barrier is therefore critical for developing brain-targeted drug delivery strategies. Although phage display screening coupled with high-throughput sequencing (HTS) enables large-scale peptide profiling, its interpretation often depends on relative abundance changes that are susceptible to amplification bias and model variability. In this dissertation, I developed a reproducibility-centered analytical framework that integrates statistical normalization with biological interpretation to systematically identify BBB-penetrating peptide candidates from both in vitro and in vivo datasets. The quantitative approach was first constructed using iterative biopanning on transwell and microfluidic BBB-chip models, allowing precise tracking of peptide enrichment and exclusion of nonspecific amplification. Functionally, top-ranked peptides from the BBB-chip model exhibited approximately three-fold higher permeability than those from the transwell, confirming that sequencing-based enrichment corresponded to genuine transport capacity. Transcriptomic analyses further demonstrated that endothelial cells under microfluidic shear expressed tight-junction and glycocalyx-related genes resembling in vivo brain endothelium, linking physiological fidelity of the model to selection accuracy. Building upon this in vitro foundation, the analytical framework was expanded to multi-set in vivo phage display datasets. Normalization and reproducibility-based filtering were applied to correct compositional bias and identify peptides consistently enriched across independent biological replicates. These reproducibility-guided criteria revealed high-confidence peptide sequences likely to represent BBB-penetrating ligands rather than stochastic artifacts. While the analyses established statistically reproducible patterns, further experimental validation is required to confirm their biological relevance. Collectively, this work establishes a data-centric and reproducibility-guided framework for BBB peptide discovery that unites biological fidelity with statistical reliability. By redefining high-throughput phage display analysis as a reproducible and physiologically grounded process, this dissertation provides a methodological and conceptual foundation for future data-driven development of brain-targeted delivery systems.