Recent Posts

Benchmarking Rcpp code with RcppClock

Timing Rcpp code is challenging. Latency between R and Rcpp, along with associated C++ library to provide an R-friendly benchmarking experience. type conversions, can confound results. The RcppClock R package wraps the std::chrono

Last updated on Nov 1, 2021 Rcpp, methods

Linked NMF for Signal Source Separation

Non-negative matrix factorization is a useful method for additive decomposition of signal within a dataset. However, NMF of concatenated datasets does not cleanly resolve batch effects, sources of heterogeneity, and common signal. Linked NMF implicitly couples independent factorizations of multiple datasets to recover models describing shared and unique signal.

Last updated on Oct 25, 2021 NMF, methods

Learning Optimal NMF Models from Random Restarts

Finding the best discoverable solution for a non-negative matrix factorization from a random initialization requires multiple random restarts. NNDSVD has previously been proposed as a “head-start” for NMF, but I show that it is not always a head start, and can be a dangerous local minima. I further explore the use of random uniform or random gaussian models for NMF initialization.

Last updated on Oct 20, 2021 NMF, methods

Learning Optimal NMF Models from Random Restarts

Statistical properties of L1- and L2-regularized NMF

L1- and L2-regularized non-negative matrix factorizations have special properties. Here I show how L1 is a sparsifying regularization that promotes a k-means clustering-like model, while L2 is a densifying regularization that promotes convergence of all factors towards the first singular vector.

Last updated on Oct 18, 2021 NMF, methods

Statistical properties of L1- and L2-regularized NMF

Cross-validation for NMF rank determination

In this post I review four distinctly different methods for cross-validation of NMF, each with strengths and weaknesses for different applications, and discuss how to use these methods effectively.

Last updated on Oct 18, 2021 NMF, methods

Annotating NMF factors with sample metadata

Use the Hawaii Birds dataset to show how to annotate factors in NMF models using metadata, with some interesting UMAP visualizations and applications to ecology along the way.

Last updated on Oct 12, 2021 NMF, annotation

Annotating NMF factors with sample metadata

Integrating Heterogenous Samples with NMF

DNA methylation signatures in patients with Acute Myelogenous Leukemia are compared to healthy cell types to determine cell-of-origin. This is a great example of why source separation is necessary to integrate mixed or heterogenous signals. By using subsets of NMF factors, similarities between these signals are exposed.

Last updated on Oct 12, 2021 NMF, integration

Integrating Heterogenous Samples with NMF

Zach DeBruine

Assistant Professor of Bioinformatics

Grand Valley State University

About

My name is Zach DeBruine and I am an Assistant Professor of Bioinformatics at Grand Valley State University and Adjunct Professor at the Van Andel Institute. I work in the Department of Computing and the Applied Computing Institute where my lab develops methods for analysis of single-cell experiments.

I developed the general-purpose Rcpp Machine Learning Library (RcppML) R package around my implementation of Non-negative Matrix Factorization (NMF), which is the world’s fastest and most flexible implementation of the method. This blog is my sandbox for trying new things, sharing stuff I’ve learned, and testing stuff I’ve built.

This blog is built with blogdown and Hugo, and deployed using Netlify.

Interests

Non-negative Matrix Factorization
Single-cell Analysis
Machine Learning

Education

Postdoctoral fellowship in Bioinformatics, 2022

Van Andel Institute
PhD in Biochemistry and Bioinformatics, 2021

Van Andel Institute
BSc in Biochemistry and Molecular Biology, 2015

Hope College