Boosting predictive models and augmenting patient data with relevant genomic and pathway information.

Area of Science:

• Oncology
• Bioinformatics
• Machine Learning

Background:

• Low-stage lung cancer recurrence is unpredictable, challenging personalized treatment.
• Current relapse prediction models lack comprehensive genetic data for accuracy.
• Early relapse identification is crucial for improving patient outcomes in lung cancer.

Purpose of the Study:

• To refine machine learning models for precise relapse prediction in early-stage non-small cell lung cancer.
• To integrate specific genetic information, such as pathway scores, into clinical data for enhanced prediction.
• To address the scarcity of genetic data by employing imputation techniques.

Main Methods:

• Leveraged The Cancer Genome Atlas (TCGA) for genetic data imputation.
• Integrated imputed pathway scores with clinical data from the Cancer Long Survivor Artificial Intelligence Follow-up (CLARIFY) project.
• Trained machine learning models on enriched knowledge graph data, including pathway score imputation triples.

Main Results:

• Achieved 82% precision and 91% specificity in predicting relapse on a held-out test set.
• Demonstrated significant strides in relapse prediction by integrating imputed pathway scores with clinical data.
• Developed machine learning models capable of enhancing prognostic capabilities.

Conclusions:

• Integrating imputed genetic pathway scores significantly improves relapse prediction accuracy in early-stage non-small cell lung cancer.
• Machine learning models trained on enriched data show promise as supplementary tools to TNM classification.
• This approach offers improved prognostic capabilities, potentially elevating patient outcomes through better relapse anticipation.