DECONbench

DECONbench benchmarks computational methods to quantify cell-type heterogeneity in cancer using standardized benchmark datasets and performance metrics.

Key Features:

• Benchmark Datasets: Provides standardized benchmark datasets for evaluating deconvolution and cell-type composition algorithms in tumor samples.
• Computational Methods Evaluation: Enables systematic comparison of diverse computational deconvolution methods for tumor heterogeneity quantification.
• Submission of New Methods: Accepts submission of new deconvolution methods for benchmarking against established datasets and metrics.
• Performance Evaluation Metrics: Applies quantitative performance metrics to assess the accuracy and efficacy of deconvolution and heterogeneity estimation methods.

Scientific Applications:

• Methodological comparison and validation: Facilitates methodological comparison and validation of deconvolution algorithms used in cancer research.
• Modeling cancer progression: Supports development of models of cancer progression through standardized quantification of tumor cell-type composition.
• Therapeutic target identification and response prediction: Aids identification of potential therapeutic targets and prediction of treatment responses by assessing tumor cell-type composition.

Methodology:

Implemented on the open-source Codalab competition platform (https://competitions.codalab.org/competitions/23660) with additional resources at https://cancer-heterogeneity.github.io/deconbench.html.