CCC-GPU wouldn't exist without the impressive work of @haoyu-zc.bsky.social, author and main developer.
From method to high-performance CUDA implementation 💻🔥
Try CCC-GPU: github.com/pivlab/ccc-gpu
You can read more on the @cubiomedinfo.bsky.social web: news.cuanschutz.edu/dbmi/cluster...
n/n
CCC-GPU reveals patterns that standard correlations miss, including relationships present only in subsets of samples, such as sex-specific gene expression effects.
4/n
The original CCC was powerful but computationally expensive.
CCC-GPU changes that:
⚡ ~37× speedup in GTEx-scale data
⚡ Whole multi-tissue analysis in hours instead of ~2 weeks
This makes large-scale nonlinear correlation analysis practical.
3/n
Popular correlation methods focus on linear or monotonic patterns.
CCC is different:
→ It uses clustering
→ Captures linear + nonlinear relationships
→ Works with mixed data types (you can correlate with metadata, if available)
2/n
Our CCC-GPU paper is published in Bioinformatics (Oxford)! 🚀
CCC is a not-only-linear correlation coefficient that uncovers complex gene-gene relationships beyond linear patterns. Now it runs on GPUs!
doi.org/10.1093/bioi...
Huge credit to @haoyu-zc.bsky.social!
1/n
a set of hexagonal logos (hexbin style display) displaying software tools. All logos are partially blurred and faded besides the CCC logo
Using a familiar correlational coefficient like Spearman’s or Pearson’s on molecular data may miss some complex patterns that exist.
To overcome this, researchers at @cuanschutz.bsky.social including @miltondp.bsky.social developed CCC—a new statistical framework ➡️ https://cudbmi.info/ccc-1
With software engineers from @cubiomedinfo.bsky.social, we developed Manugen-AI, an agentic AI system to automate the mechanical tasks of writing a scientific article #adkhackathon Check it out!
Devpost: devpost.com/software/aca...
Code: github.com/pivlab/manug...
Medium: medium.com/@miltondp/ma...