MIDAS’ reproducibility documentation

—An instruction for reproducing the manuscript results.

MIDAS turns mosaic data into imputed and batch-corrected data to support single-cell multimodal analysis.

Contents:

• Preparation
  • Developing and test environment
  • Quality control for individual datasets (27 batches in total)
• Mosaic integration
  • Generating training data for MIDAS
  • Training MIDAS
  • Inferring latent variables, and generating imputed and batch corrected counts
  • Evaluation
• Atlas construction
  • Generating training data for MIDAS
  • Training MIDAS
  • Inferring latent variables, and generating imputed and batch corrected counts
  • Evaluation
• Knowledge transfer
  • Reference construction
  • Mosaic integration with fewer modalities
  • Model transfer
  • Label transfer