iAntiSplodge

iAntiSplodge deconvolves mixed mRNA profiles from spatial transcriptomics (ST) data to resolve cell-type composition using neural-network models trained on synthetic ST profiles derived from single-cell RNA datasets.

Key Features:

• Neural Network Architecture: Employs a simple feed-forward neural network for deconvolution of ST spot profiles.
• Synthetic ST Profile Utilization: Generates and uses synthetic spatial transcriptomics profiles derived from real single-cell RNA datasets for training and inference.
• Data Integration: Incorporates single-cell RNA data from matching tissues to inform deconvolution models.
• Performance Validation: Validated against state-of-the-art tools demonstrating improved accuracy and computational efficiency.

Scientific Applications:

• Human heart transcriptomics: Applied to deconvolute human heart ST data across time points to analyze temporal cell-type distributions.
• Mouse brain analysis: Applied to mouse brain ST data with spot patterns aligning to anatomical structures such as the hippocampus.

Methodology:

Incorporates single-cell RNA data from similar tissues, trains a simple feed-forward neural network on synthetic ST profiles derived from single-cell datasets, and validates performance against state-of-the-art tools.