Collection of Lymphoma Somatic Mutations from 57 Published NGS Studies for Comparative Analysis

Examples of Deep Machine Learning on the Entire LymphoDB database

    Case Study I: Identifing Lymphome Subtype Specific Mutation Patterns and Their Associated Pathways
    The heat map view on the unsupervised clustering of 13 lymphoma subtypes and their mutation patterns. The color red represents mutations. The result shows that the lymphoma subtypes have common and/or unique mutations patterns. Notably, some subtypes have very distinctive mutation patterns, which suggests that they have very different oncogenesis mechanism. For example, the unique mutation set of MZL heavily hits RAS signaling pathway. The example of the LymphoDB + certain machine learning approaches could advance the lymphoma research by identifying subtype specific mutation patterns and associated pathways. Such subtype specific patterns could not be previously done without the large integration efforts of LymphoDB.

Case Study II:
Mutation-based Diagnosis Implication
    The close subtypes of DLBCL and BL can be discriminated by their mutation profiles. Each dot represents one of 50 patients selected from LymphoDB, where DLBCL is represented in purple and BL in blue. Based on their mutation profiles, the mutliscaling analysis can systematically distinguish them into the right groups without the input of any human knowledge. Currently, morphological and biomarker based diagnosis has a high rate of failure in correctly seperating DLBCL vs. BL, thereafter leading to the incorrect treatment protocol for more toxic BL patients. This analysis suggests the promising clincial potential of mutation based diagnosis of DLBCL vs. BL. LymphoDB's mutation information regarding thousands of lymphoma patients, provides important resources for scientists to explore these types of analysis.

Dr Louis Staudt' Laboratory

LymphoDB was developed by several high school students as an internship project in xxxx Research Institute

The project is being under a Scientific Competition so that all IDs, names, and logos of students, mentors and institute are anonymous throughout the web site as required

To hide our institute web address, this is a clone of our official web site in a third party server for a Scientific Competition