Sozzani Lab - Research Assistant (Deep Learning)
Location: Raleigh, NC
Duration: Sept 2023 - May 2024
Protein Sequence Classification with Deep Learning
- Developed a deep learning architecture combining CNN, Attention, and LSTM to classify protein families and annotate protein sequences.
- Enabled functional inference of signaling networks, facilitating understanding of gene regulation and protein interactions.
- Achieved an F1 score of 96% on the classification task using High Performance Computing (HPC) for large-scale sequence training.
- Research accepted for publication in the Nature Journal.
Clustering of Arabidopsis Classes with Variational AutoEncoders
- Designed and implemented a dimensionality reduction and clustering pipeline using AutoEncoders combined with K-Means.
- Identified distinct subtypes of Arabidopsis plant phenotypes for deeper understanding of genetic variations.
- Built and trained models using PyTorch and tracked experiments and hyperparameters with MLflow.
Graph Convolution Networks for GRN Inference
- Proposed a novel Graph Convolutional Network (GCN) with attention mechanisms to infer Gene Regulatory Networks (GRNs).
- Used protein interaction graphs to identify functional regions and cell type changes in Arabidopsis plants.
- Enabled spatial understanding of protein roles across different developmental stages in plant biology.