Skip to content

Multiomics data integration with quantile matrix factorization

License

Notifications You must be signed in to change notification settings

jpvert/qmf-genomics

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Multiomics data integration with quantile matrix factorization

Quantile matrix factorization (QMF) is a technique to approximate a matrix by a low-rank matrix followed by row-wise monotonic transform. Here we test its ability to identify factors associated with survival in bulk multi-omics TCGA data, and compare it to plain nonnegative matrix factorization (NMF), following the MOMIX benchmark proposed by Cantini et al. (2020).

Data provided

We provide the pre-computed matrix factorization by NMF and QMF in the data/factorizations/ directory. Note that we provide two factorizations for each method and each cancer, obtained by running the optimization twice with two different random initializations. We also provide the survival information for each cancer, as provided by the MOMIX benchmark, in the data/cancer/ directory.

Run the benchmark

To run the benchmark on the NMF and QMF factorizations, just run the Jupyter notebook (adapted from the MOMIX one):

jupyter-notebook script/runbenchmark.ipynb

It should produce a series of plots, as in the QMF paper cited below.

Citation

Cuturi, M., Teboul, O., Niles-Weed, J., and Vert, J.-P. (2020), “Supervised Quantile Normalization for Low-rank Matrix Approximation,” in Thirty-seventh International Conference on Machine Learning (ICML 2020). To appear. PDF

About

Multiomics data integration with quantile matrix factorization

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published