Publications

Structured and Target-Specific Development of Cortico-Cortical Connectivity in the Mouse Visual Cortex

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

MOAST: Mechanism of Action Similarity Tool

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.

Massively parallel, single-molecule assessment of synthetic fidelity and drug-like properties in a DNA-encoded library

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.

BPS2025 - Computational design of NaV1.8 sodium channel inhibitors as novel non-addictive treatments for pain management

Structural modeling of ion channels using AlphaFold2, RoseTTAFold2, and ESMFold

Ion channels play key roles in human physiology and are important targets in drug discovery. The atomic-scale structures of ion channels provide invaluable insights into a fundamental understanding of the molecular mechanisms of channel gating and modulation. Recent breakthroughs in deep learning-based computational methods, such as AlphaFold, RoseTTAFold, and ESMFold have transformed research in protein structure prediction and design. We review the application of AlphaFold, RoseTTAFold, and ESMFold to structural modeling of ion channels using representative voltage-gated ion channels, including human voltage-gated sodium (NaV) channel - NaV1.8, human voltage-gated calcium (CaV) channel – CaV1.1, and human voltage-gated potassium (KV) channel – KV1.3. We compared AlphaFold, RoseTTAFold, and ESMFold structural models of NaV1.8, CaV1.1, and KV1.3 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.

Computational design of binders targeting the VSDIV from NaV1.7 sodium channel