wenda_gpu

wenda_gpu implements GPU-accelerated weighted elastic net domain adaptation for genomic predictive modeling, enabling efficient transfer learning and prediction of cancer mutation status from limited sample sizes.

Key Features:

• Accelerated computation: Integrates GPyTorch to train models on a single GPU, enabling processing of genome-sized genomic datasets within hours.
• Comparable performance: Maintains the accuracy and reliability of the original weighted elastic net domain adaptation (wenda) while improving computational efficiency.
• Enhanced predictive capabilities: Improves prediction in small-sample scenarios and yields superior cancer mutation status prediction compared to standard elastic net methods.

Scientific Applications:

• Mutation-status prediction: Predicts cancer mutation status from genomic datasets with limited target samples.
• Personalized treatment planning: Supports development of genomic-informed personalized treatment strategies by enabling domain-adapted predictive models.
• Cancer genomics research: Facilitates research into the genetic underpinnings of various cancers via rapid domain adaptation of models across datasets.

Methodology:

Implements weighted elastic net domain adaptation and uses GPyTorch for GPU-accelerated model training on large-scale genomic datasets.