UniFORM

UniFORM normalizes multiplex tissue imaging (MTI) data at both the feature and pixel levels using an automated rigid landmark registration method that removes technical variation while preserving biologically relevant expression patterns.

Key Features:

• Non-parametric normalization: UniFORM operates without prior distributional assumptions and handles both unimodal and bimodal patterns.
• Automated rigid landmark registration: The method is tailored to the distributional characteristics of MTI and aligns biologically invariant negative populations.
• Positive population preservation: Technical variation is removed while tissue-specific expression patterns in positive populations are maintained.
• Guided fine-tuning mode: An optional mode is available for complex or heterogeneous datasets.
• Scalable design: The pipeline supports broad applications for MTI data normalization beyond fluorescence-based MTI.

Scientific Applications:

• Batch effect mitigation: UniFORM outperforms existing methods in reducing batch effects across three MTI platforms, as measured by improved marker distribution alignment, enhanced k-nearest neighbor batch effect test (kBET) scores, and improved silhouette scores.
• Downstream clustering and visualization: Normalized data produces more coherent uniform manifold approximation and projection (UMAP) visualizations and Leiden clustering results.
• Cross-platform MTI analysis: The method enables accurate and biologically meaningful interpretations across different multiplex tissue imaging platforms.

Methodology:

UniFORM employs automated rigid landmark registration tailored to the distributional characteristics of MTI, aligning biologically invariant negative populations to remove technical variation while preserving tissue-specific expression patterns in positive populations.