Research Skills

High-dimensional & multi-modal data analysis

• In my Ph.D., I analyzed high-dimensional single-cell datasets using sparse cell x gene matrices with thousands of individual features and up to tens of millions of unique cells:

	Gene_1	Gene_2	⋯	Gene_m
Cell_1	⋯	⋯	⋯	⋯
Cell_2	⋯	⋯	⋯	⋯
⋯	⋯	⋯	⋯	⋯
Cell_n	⋯	⋯	⋯	⋯

• These datasets contained multiple cell x feature matrices with the multi-modal profiles for each cell

• To analyze these datasets, I used unsupervised machine-learning methods such as t-SNE and Louvain clustering:

Reagor & Hudspeth, 2024, bioRxiv

---

Deep learning for causal time-series analysis

• I developed the deep-learning method DELAY to reconstruct causal gene-regulatory networks from single-cell gene-expression datasets

• Using concepts from Granger Causality, I designed DELAY to encode noisy gene-expression data as images for deep learning

Reagor, Velez-Angel & Hudspeth, 2023, PNAS Nexus

• DELAY uses a convolutional neural network to classify images as either interacting or non-interacting gene pairs

---

Analysis of large-scale networks

• To reconstruct large-scale networks, I inferred causal interactions between small clusters of genes

• I used graph theory to identify the nodes (genes) that control the network’s temporal dynamics

To validate this reconstructed network, I performed experiments on regenerating zebrafish

• Beyond DELAY, I also developed custom scripts to quantify and analyze neuronal networks