PredCID

PredCID predicts cancer driver frameshift indels by integrating gene-, DNA-, transcript-, and protein-level features to distinguish driver from passenger mutations in human cancers.

Key Features:

• Multi-level genomic feature integration: Integrates features at the gene, DNA, transcript, and protein levels to predict cancer driver frameshift indels.
• Feature selection: Selects relevant features from these biological layers to improve prediction.
• XGBoost-based classification: Employs an eXtreme Gradient Boosting (XGBoost) classifier for driver/passenger classification.
• Robustness to missing values: Mitigates missing values that arise from inconsistent transcript choices during variant annotation.
• Benchmark performance: Reported benchmarking shows improved performance over noncancer-specific methods across cross-validation and independent datasets.

Scientific Applications:

• Frameshift indel interpretation in human cancers: Distinguishes predicted cancer driver frameshift indels from passenger mutations to support cancer genomics analyses.

Methodology:

Selects features from gene-, DNA-, transcript-, and protein-level annotations and applies an XGBoost classifier for driver/passenger classification, mitigating missing values caused by inconsistent transcript selection during variant annotation.