Does this skill work with any single-cell dataset?

While specifically validated on PhenoCycler and lymph node datasets, the mathematical logic applies to any cross-modal integration using MaxFuse where modal imbalances are extreme.

Why is sqrt scaling better than linear scaling for matching ratios?

In large datasets, the pivot propagation mechanism doesn't require every protein cell in the initial batch; linear scaling causes over-sampling and noise that results in good matches being filtered away.

When should I use the Large Cell Ratio Matching skill?

Use this skill when your protein-to-RNA cell ratio exceeds 100:1 and you notice significant RNA cell loss (over 10%) after performing a MaxFuse integration.

How does GMM-guided filtering improve integration results?

Instead of assuming a fixed 20% error rate, GMM-guided filtering analyzes the actual bimodal score distribution to identify and remove only the truly poor matches.

Large Cell Ratio Matching for MaxFuse

Name: Large Cell Ratio Matching for MaxFuse
Author: smith6jt-cop

bysmith6jt-cop

数据科学与机器学习

Optimizes MaxFuse parameters to prevent RNA cell loss in datasets with extreme protein-to-RNA imbalance ratios.

关于

This skill provides specialized implementation patterns for MaxFuse integration when dealing with protein-to-RNA cell ratios exceeding 100:1. It solves the common problem of excessive cell filtering in large-scale spatial biology datasets by replacing default linear scaling with sqrt-scaled matching ratios and implementing adaptive GMM-guided pivot filtering. This approach ensures RNA cell coverage remains above 90% while maintaining high matching accuracy for PhenoCycler, CODEX, and other cross-modal spatial transcriptomics workflows.

主要功能

Adaptive GMM-guided pivot filtering based on actual score distributions
Validated workflows for datasets with 1.7M+ protein cells
Sqrt-scaled matching ratio calculation to prevent noisy over-sampling
Automated diagnostic checks for dataset modal imbalance
0 GitHub stars
Conservative propagation filtering logic to minimize cumulative data loss

使用场景

Integrating large-scale PhenoCycler protein datasets with sparse RNA-seq data
Recovering lost RNA cell coverage in multi-modal spatial biology pipelines
Optimizing MaxFuse performance for high-resolution spatial proteomics

关于

主要功能

Adaptive GMM-guided pivot filtering based on actual score distributions
Validated workflows for datasets with 1.7M+ protein cells
Sqrt-scaled matching ratio calculation to prevent noisy over-sampling
Automated diagnostic checks for dataset modal imbalance
0 GitHub stars
Conservative propagation filtering logic to minimize cumulative data loss

使用场景

Integrating large-scale PhenoCycler protein datasets with sparse RNA-seq data
Recovering lost RNA cell coverage in multi-modal spatial biology pipelines
Optimizing MaxFuse performance for high-resolution spatial proteomics