Kernel Methods for Genomic Data
- Monday, May 3, 2021 from 10:00am to 11:00am
Identifying characteristics of genomic data and annotating them is an important and challenging task in biology. As the experimental methods can be expensive, there are different computational methods applied for these tasks. Genomic data can be modeled as graphs that can carry sequential, structural, and physicochemical information within them. Due to this complex structure, some of the computational methods are unable to capture all the features in them. As a solution to this, graph kernel methods are adapted to these genomic data. Graph kernels are able to compare the similarities between graphs in multiple levels. Therefore, they are being used in multiple applications in genomics. The graphs kernels are based on different graph components such as subgraphs, walks/paths and, node and edge labels to better serve the problem. This bibliographic report summarises some of the graph kernel methods that were adapted for genomics. After analyzing these methods, it is apparent that there are more aspects to expand these methods further and make them more robust and efficient.
- Gianforte School of Computing