Skip to main content
Fig. 1 | Genome Biology

Fig. 1

From: GRiNCH: simultaneous smoothing and detection of topological units of genome organization from sparse chromatin contact count matrices with matrix factorization

Fig. 1

Overview of GRiNCH. GRiNCH applies non-negative matrix factorization (NMF) to a Hi-C or a similar high-throughput 3C matrix to find clusters of densely interacting genomic regions. NMF recovers low-dimensional factors U and V of the input matrix X that can be used to reconstruct the input matrix. As nearby genomic regions tend to interact more with each other, we regularize the factor matrices with a neighborhood graph to encourage neighboring regions to have a similar lower-dimensional representation, and subsequently belong to the same cluster. We cluster the regions by treating one of the factor matrices as a set of latent features and applying k-medoids clustering. The clusters represent topological units such as TADs. The factor matrices can be multiplied together to yield a smoothed version of the input matrix which is often sparse and noisy

Back to article page