BERMUDA

BERMUDA applies transfer learning and deep autoencoders to remove batch effects from single-cell RNA sequencing (scRNA-seq) data and preserve genuine transcriptional signals across multiple datasets.

Key Features:

• Transfer learning: Transfers information across batches to integrate multiple scRNA-seq datasets with diverse cell population compositions.
• Deep autoencoders: Uses deep autoencoder neural networks to capture complex patterns in high-dimensional scRNA-seq data and distinguish biological signals from batch effects.
• Batch effect correction: Aligns different batches to reduce technical confounding while retaining biological variability such as cell-type differences.
• Performance on benchmarks: Demonstrated superior performance in distinguishing cell types and reducing batch effects on simulated and real scRNA-seq datasets.

Scientific Applications:

• Cell lineage identification: Enhances identification and tracing of cell lineages by reducing technical variability in integrated scRNA-seq data.
• Transcriptional signal discovery: Improves detection of bona fide transcriptional signals and novel cellular subtypes across combined datasets.

Methodology:

BERMUDA employs transfer learning and deep autoencoders to remove batch effects by aligning different scRNA-seq batches and was evaluated on simulated and real scRNA-seq datasets.