Ckmeans.1d.dp - Optimal, Fast, and Reproducible Univariate Clustering

Fast, optimal, and reproducible weighted univariate clustering by dynamic programming. Four problems are solved, including univariate k-means (Wang & Song 2011) <doi:10.32614/RJ-2011-015> (Song & Zhong 2020) <doi:10.1093/bioinformatics/btaa613>, k-median, k-segments, and multi-channel weighted k-means. Dynamic programming is used to minimize the sum of (weighted) within-cluster distances using respective metrics. Its advantage over heuristic clustering in efficiency and accuracy is pronounced when there are many clusters. Multi-channel weighted k-means groups multiple univariate signals into k clusters. An auxiliary function generates histograms adaptive to patterns in data. This package provides a powerful set of tools for univariate data analysis with guaranteed optimality, efficiency, and reproducibility, useful for peak calling on temporal, spatial, and spectral data.

Last updated 1 years ago

8.46 score 19 stars 19 packages 333 scripts 5.4k downloads

OptCirClust - Circular, Periodic, or Framed Data Clustering: Fast, Optimal, and Reproducible

Fast, optimal, and reproducible clustering algorithms for circular, periodic, or framed data. The algorithms introduced here are based on a core algorithm for optimal framed clustering the authors have developed (Debnath & Song 2021) <doi:10.1109/TCBB.2021.3077573>. The runtime of these algorithms is O(K N log^2 N), where K is the number of clusters and N is the number of circular data points. On a desktop computer using a single processor core, millions of data points can be grouped into a few clusters within seconds. One can apply the algorithms to characterize events along circular DNA molecules, circular RNA molecules, and circular genomes of bacteria, chloroplast, and mitochondria. One can also cluster climate data along any given longitude or latitude. Periodic data clustering can be formulated as circular clustering. The algorithms offer a general high-performance solution to circular, periodic, or framed data clustering.

Last updated 3 years ago

4.42 score 2 packages 22 scripts 165 downloads

FunChisq - Model-Free Functional Chi-Squared and Exact Tests

Statistical hypothesis testing methods for inferring model-free functional dependency using asymptotic chi-squared or exact distributions. Functional test statistics are asymmetric and functionally optimal, unique from other related statistics. Tests in this package reveal evidence for causality based on the causality-by- functionality principle. They include asymptotic functional chi-squared tests (Zhang & Song 2013) <doi:10.48550/arXiv.1311.2707>, an adapted functional chi-squared test (Kumar & Song 2022) <doi:10.1093/bioinformatics/btac206>, and an exact functional test (Zhong & Song 2019) <doi:10.1109/TCBB.2018.2809743> (Nguyen et al. 2020) <doi:10.24963/ijcai.2020/372>. The normalized functional chi-squared test was used by Best Performer 'NMSUSongLab' in HPN-DREAM (DREAM8) Breast Cancer Network Inference Challenges (Hill et al. 2016) <doi:10.1038/nmeth.3773>. A function index (Zhong & Song 2019) <doi:10.1186/s12920-019-0565-9> (Kumar et al. 2018) <doi:10.1109/BIBM.2018.8621502> derived from the functional test statistic offers a new effect size measure for the strength of functional dependency, a better alternative to conditional entropy in many aspects. For continuous data, these tests offer an advantage over regression analysis when a parametric functional form cannot be assumed; for categorical data, they provide a novel means to assess directional dependency not possible with symmetrical Pearson's chi-squared or Fisher's exact tests.

Last updated 7 months ago

4.37 score 29 scripts 484 downloads

GridOnClusters - Cluster-Preserving Multivariate Joint Grid Discretization

Discretize multivariate continuous data using a grid that captures the joint distribution via preserving clusters in the original data (Wang et al. 2020) <doi:10.1145/3388440.3412415>. Joint grid discretization is applicable as a data transformation step to prepare data for model-free inference of association, function, or causality.

Last updated 7 months ago

2.70 score 3 scripts 708 downloads

TreeDimensionTest - Trajectory Presence and Heterogeneity in Multivariate Data

Testing for trajectory presence and heterogeneity on multivariate data. Two statistical methods (Tenha & Song 2022) <doi:10.1371/journal.pcbi.1009829> are implemented. The tree dimension test quantifies the statistical evidence for trajectory presence. The subset specificity measure summarizes pattern heterogeneity using the minimum subtree cover. There is no user tunable parameters for either method. Examples are included to illustrate how to use the methods on single-cell data for studying gene and pathway expression dynamics and pathway expression specificity.

Last updated 3 years ago

2.70 score 1 scripts 117 downloads

DiffXTables - Pattern Analysis Across Contingency Tables

Statistical hypothesis testing of pattern heterogeneity via differences in underlying distributions across multiple contingency tables. Five tests are included: the comparative chi-squared test (Song et al. 2014) <doi:10.1093/nar/gku086> (Zhang et al. 2015) <doi:10.1093/nar/gkv358>, the Sharma-Song test (Sharma et al. 2021) <doi:10.1093/bioinformatics/btab240>, the heterogeneity test, the marginal-change test (Sharma et al. 2020) <doi:10.1145/3388440.3412485>, and the strength test (Sharma et al. 2020) <doi:10.1145/3388440.3412485>. Under the null hypothesis that row and column variables are statistically independent and joint distributions are equal, their test statistics all follow an asymptotically chi-squared distribution. A comprehensive type analysis categorizes the relation among the contingency tables into type null, 0, 1, and 2 (Sharma et al. 2020) <doi:10.1145/3388440.3412485>. They can identify heterogeneous patterns that differ in either the first order (marginal) or the second order (differential departure from independence). Second-order differences reveal more fundamental changes than first-order differences across heterogeneous patterns.

Last updated 4 years ago

2.70 score 2 scripts 214 downloads

CircularSilhouette - Fast Silhouette on Circular or Linear Data Clusters

Calculating silhouette information for clusters on circular or linear data using fast algorithms. These algorithms run in linear time on sorted data, in contrast to quadratic time by the definition of silhouette. When used together with the fast and optimal circular clustering method FOCC (Debnath & Song 2021) <doi:10.1109/TCBB.2021.3077573> implemented in R package 'OptCirClust', circular silhouette can be maximized to find the optimal number of circular clusters; it can also be used to estimate the period of noisy periodical data.

Last updated 3 years ago

2.48 score 1 packages 3 scripts 224 downloads