Computational Peptide Drug Design & Neuroscience
Research in computational peptide drug design and developmental neuroscience using machine learning, evolutionary and stochastic optimization, and statistical and graph-based methods research at UC Santa Cruz and UC Davis.
- 2026 BPS2026 – De novo design of ion channel modulators using deep learning for therapeutic applications Biophysical Journal · 10.1016/J.BPJ.2025.11.2172
- 2025-11-05 Structured and Target-Specific Development of Cortico-Cortical Connectivity in the Mouse Visual Cortex Preprint · 10.1101/2025.11.03.686379 The mammalian cortex exhibits highly stereotyped long-range connectivity, yet the developmental principles that specify precise cortico-cortical projection patterns remain poorly defined. Two dominant models propose that target specificity arises either from early inter-regional exuberant outgrowth followed by pruning, or through initially directed axonal targeting. To resolve this, we systematically mapped the postnatal development of V1 cortico-cortical projection neurons (CCPNs) to eleven higher visual areas (HVAs) in mice using rapid and complementary retrograde, anterograde, and single-cell tracing methods. We found that V1→HVA connectivity develops via spatiotemporally staggered axon extension and pruning programs, aligned with target position along the medial-lateral axis. Reciprocal HVA→V1 feedback emerges concurrently and is refined over time, yielding gradually aligned bidirectional connectivity. Notably, both multiplexed retrograde tracing and MAPseq-based single-cell profiling revealed that individual V1 neurons initialize and retain specific projection motifs with limited variation over development, arguing against global exuberance followed by selective, inter-areal pruning. Instead, our findings support a directed guidance model, in which distinct V1 CCPN subtypes establish selective projection patterns early, followed by local, target-dependent refinement. This structured yet heterogeneous developmental strategy provides an anatomical framework for how precise long-range cortical networks emerge. HIGHLIGHTS V1→HVA connections form via directed axonal targeting, establishing motifs early with little variation Medial targets are innervated earlier and refine gradually, lateral targets later and rapidly Feedforward and feedback V1-HVA circuits emerge concurrently Bidirectional like-to-like V1–HVA connectivity refines across development
- 2025-09-19 MOAST: Mechanism of Action Similarity Tool Preprint · 10.1101/2025.09.15.676411 Determining the mechanism of action (MOA) for natural products remains a significant bottleneck in drug discovery, particularly for researchers with limited computational resources or small compound libraries. Traditional approaches require screening large numbers of annotated compounds alongside unknowns, which is cost-prohibitive, or depend on complex machine learning models that need substantial computational resources and large datasets. Here, we present a dissertation chapter excerpt: MOAST (Mechanism of Action Similarity Tool), a BLAST-inspired computational workflow that addresses these limitations by providing rapid MOA hypotheses for newly screened compounds. This chapter investigates two complementary approaches: a kernel density estimation (KDE) method providing statistical significance measures and E-values for MOA class membership, and a CatBoost machine learning classifier for multi-class prediction with ranked outputs. Using cytological profiling data from HeLa and A549 cell lines, MOAST achieved 22% accuracy for the top 5 predictions among ∼ 300 MOA classes, with the CatBoost classifier reaching 10% balanced accuracy—significantly better than the ∼ 3% reported in literature. The tool suggests a 0.8 prediction probability threshold for trustworthy results and demonstrates robust performance across multiple feature reduction strategies. MOAST provides a practical, accessible solution that bridges traditional phenotypic screening and modern computational approaches, making MOA determination feasible for researchers with limited resources while maintaining statistical rigor and interpretability. 8 refs
- 2025-07-14 Massively parallel, single-molecule assessment of synthetic fidelity and drug-like properties in a DNA-encoded library Preprint · 10.26434/chemrxiv-2025-rlwvx DNA-encoded libraries (DELs) have emerged as a promising drug discovery strategy, but successful translation of hits is often impeded by synthetic inefficiency and enrichment of poorly permeable compounds. Here we introduce a novel sequencing-based separation strategy, LC-seq, that simultaneously evaluates synthetic fidelity and permeability-relevant lipophilicity for individual DNA-encoded library members. Using a 120,000-member peptide library, we mapped reaction efficiency across all synthetic cycles and identified structure-reactivity trends. The on-DNA lipophilicities for resynthesized library members correlate strongly with their off-DNA lipophilicities and passive permeability in artificial membranes and MDCK cells. This approach enables direct assessment of compound quality and drug-like properties at unprecedented scale, potentially transforming DEL-based drug discovery. 1 cited
- 2025 Computational design of NaV1.8 sodium channel inhibitors as novel non-addictive treatments for pain management BIOPHYSICAL JOURNAL
- 2024 Computational design of binders targeting the VSDIV from NaV1.7 sodium channel BIOPHYSICAL JOURNAL
- 2024 Structural modeling of ion channels using AlphaFold2, RoseTTAFold2, and ESMFold CHANNELS · 10.1080/19336950.2024.2325032 1.3 with corresponding cryo-EM structures to assess details of their similarities and differences. Our findings shed light on the strengths and limitations of the current state-of-the-art deep learning-based computational methods for modeling ion channel structures, offering valuable insights to guide their future applications for ion channel research. 29 cited 45 refs