Files and links (1)
PDFOpen Access (License Unspecified), Open Access
Thesis
Cell-adhesion molecule expression prediction from transcription factor expression patterns
Master of Science (M.S.), Drexel University
May 2025
DOI:
https://doi.org/10.17918/00010958
Abstract
Understanding how neurons acquire specific connectivity patterns during brain development is a central challenge in neuroscience. While single-cell RNA sequencing (scRNA-seq) has enabled unprecedented resolution of transcriptional programs, most computational approaches focus on cell type classification or pseudotime inference rather than modeling regulatory logic at the gene level. In this study, we present a predictive framework for inferring the expression of cell-adhesion molecules (CAMs)--key mediators of synaptic specificity--from transcription factor (TF) expression patterns in Drosophila melanogaster. We begin by motivating the biological premise: TFs regulate CAMs, and CAMs encode partner identity. In Chapter 1, we position CAM prediction as a means of decoding neuronal wiring programs. In Chapter 2, we formulate the task as a multi-output regression problem and benchmark four machine learning models--linear regression, XGBoost, shallow neural networks, and deep neural networks (DNNs). DNNs achieve the best performance, accurately capturing the nonlinear and combinatorial nature of TF-CAM regulation. Using SHAP values, we identify cell type-specific regulatory influences, including the transcription factor broad (br) in LPLC1 neurons. Chapter 3 validates the model's generalization through perturbation. We simulate br overexpression in LPLC2 neurons and compare predicted CAM profiles to those from Perturb-seq experiments, observing strong alignment (R² = 0.86). Finally, Chapter 4 explores biological implications and extensions, proposing integration of additional modalities such as scATAC-seq and enhancer-promoter maps to improve constraint and interpretability. This work bridges data-driven modeling with experimental neurogenetics, offering a scalable approach for learning gene-level wiring logic from transcriptomic data.
Metrics
32 File views/ downloads
11 Record Views
Details
- Title
- Cell-adhesion molecule expression prediction from transcription factor expression patterns
- Creators
- Kieran James Lynch
- Contributors
- Gail L. Rosen (Advisor)
- Awarding Institution
- Drexel University
- Degree Awarded
- Master of Science (M.S.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Number of pages
- x, 39 pages
- Resource Type
- Thesis
- Language
- English
- Academic Unit
- College of Engineering (1970-2026); Electrical (and Computer) Engineering [Historical]; Drexel University
- Other Identifier
- 991022054939904721