A Crowdsourcing Approach to Developing and Assessing Prediction Algorithms for AML Prognosis.

Area of Science:

• Hematology
• Proteomics
• Computational Biology
• Cancer Prognostics

Background:

• Acute Myeloid Leukemia (AML) is a fatal hematological cancer with significant genetic heterogeneity.
• Accurate prognosis and treatment selection for AML are challenging due to this heterogeneity.
• Clinical proteomics data offers potential for improved AML prognosis, but quantitative methods are lacking.

Purpose of the Study:

• To promote the development of quantitative methods for Acute Myeloid Leukemia prognosis prediction.
• To identify the most accurate and robust models for predicting patient response to therapy, remission duration, and overall survival in AML.
• To investigate the utility of proteomics data in predicting therapeutic response and identify key predictive proteins.

Main Methods:

• The study utilized the DREAM 9 Acute Myeloid Prediction Outcome Prediction Challenge (AML-OPC), a crowdsourcing initiative.
• 31 computational models were submitted to predict AML patient outcomes based on clinical and proteomics data.
• Model performance was evaluated on predicting patient response to therapy, remission duration, and overall survival.

Main Results:

• The challenge successfully identified accurate and robust models for AML prognosis.
• Patients classified as resistant to therapy were more challenging to predict than responsive patients across all models.
• The top-performing models heavily utilized proteomics data, demonstrating its value in predicting therapeutic response and survival.

Conclusions:

• Quantitative proteomics data can significantly enhance the accuracy of Acute Myeloid Leukemia prognosis.
• The development of robust predictive models is crucial for improving patient outcomes in AML.
• Specific signaling proteins were identified as valuable predictors of patient therapeutic response in AML.