Orchid: a novel management, annotation and machine learning framework for analyzing cancer mutations

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Growing whole-genome tumor sequence and biological annotation datasets necessitate efficient data management and analysis software.
• Increasingly sophisticated data stores, execution environments, and machine learning algorithms require integrated functionality across frameworks.

Purpose of the Study:

• To present Orchid, a Python-based software package for managing, annotating, and performing machine learning on cancer mutations.
• To demonstrate Orchid's efficiency in handling large-scale mutation data and its capability in classifying tumor types.

Main Methods:

• Development of Orchid, a Python software package utilizing parallel workflow execution, in-memory database storage, and machine learning analytics.
• Implementation of Orchid using Python 2.7 with MySQL or MemSQL databases, optionally requiring Groovy 2.4.5 for parallel execution.
• Application of a random forest classifier with 339 features to distinguish tissue of origin across 12 tumor types.

Main Results:

• Orchid efficiently manages millions of mutations and hundreds of features.
• The software successfully distinguished the tissue of origin in 12 tumor types.
• Demonstrated the utility of machine learning for cancer mutation analysis and classification.

Conclusions:

• Orchid provides an efficient and user-friendly solution for cancer mutation data management and analysis.
• The software facilitates the integration of data management, annotation, and machine learning functionalities.
• Orchid aids in advancing basic science and clinical applications through improved cancer mutation analysis.