Integrating AI and Network Science to Analyze Biological Data and Solve Constrained Optimization

Dissertation proposal
Hybrid
Past Talk
Wan He
PhD Student, Network Science Institute
Tue, Jan 21, 2025
2:00 PM UTC
Tue, Jan 21, 2025
2:00 PM UTC
In-person
4 Thomas More St
London E1W 1YW, UK
The Roux Institute
Room
100 Fore Street
Portland, ME 04101
Network Science Institute
2nd floor
Network Science Institute
11th floor
177 Huntington Ave
Boston, MA 02115
Room
58 St Katharine's Way
London E1W 1LP, UK

Talk recording

Advances in next generation sequencing technology are producing large amounts of biological data. In this dissertation, I focus on the integration of AI and network science to analyze genomics and transcriptomics data. Specifically, I answer the following questions: (1) How can we compare a large collection of long genome sequences? (2) How can we use hypergraphs to improve clustering of single-cell RNA sequencing data? (3) Can representation learning on single-cell RNA sequencing data create co-expression networks with a higher signal-to-noise ratio? For the first question, I find that misclassifications from AI models provide insights for comparative genome analysis. In particular, misclassification likelihoods reveal (spatial) associations between genome ensembles. For the second question, I identify inflated signals in co-expression networks due to data sparsity; and show that using hypergraphs and co-expression networks together with a memory mechanism outperforms established methods, especially for weakly modular data. The third question is ongoing work. However, preliminary results show that embeddings from representation learning methods produce networks with less noise, which in turn leads to more distinct communities. I also investigate a related research question: How can we use hypergraphs to solve constrained optimization problems? I find that for resource allocation problems, optimizing for the algebraic connectivity of the hypergraph leads to robust and resilient solutions.

About the speaker
Wan is a PhD student in Network Science. She works with Professors Tina Eliassi-Rad and Samuel Scarpino. Her work is at the intersection of deep learning, network science and bioinformatics. Prior to joining the Network Science Institute, she worked on community detection using stochastic block models and was part of the Global Mean Surface Temperature (GMST) Reconstruction Team working on climate change projects. She received her B.S. in Mathematics from Imperial College London.
Share this page:
Jan 21, 2025