Pathway analysis helps us understand the biology behind different traits by looking beyond individual genes, proteins, or other molecules. This type of analysis is crucial, which is why over 100 methods have been developed for it. However, these methods often have biases that can lead to incorrect and inconsistent results. For instance, cancer-related pathways might show up as important in studies unrelated to cancer. Additionally, most methods focus on just one type of biological data and don’t consider the complex interactions between different types of data.
Our project aims to fix these issues in two main ways:
1. Correcting Biases: We will start by addressing biases in 20 well-known analysis methods that are widely used and highly cited. We’ll develop techniques to reduce or eliminate these biases for each method and disease.
2. Integrating Multi-Omics Data: We will create a new approach that combines different types of biological data to analyze pathways. This will involve redesigning existing pathways to include more detailed interactions based on current databases and expert knowledge. We’ll also develop a model that can learn these interactions from data. These improved pathways can then be used with the 20 existing methods to analyze complex biological data.
Finally, we will combine the corrected methods and our new multi-omics approach using a technique we developed called Consensus Pathway Analysis. This will help us identify important pathways and subnetworks for each disease, which can then be used to predict health risks and patient outcomes.