README.md

In rlebron-bioinfo/gnlearn: Genetic Network Learning

gnlearn

Genetic Network Learning (gnlearn) is an R package for structural learning of transcriptional regulatory networks from single-cell datasets. Learned transcriptional regulatory networks can be obtained as graphs (in graph or igraph R packages formats), adjacent matrices or lists of edges.

Constraint-Based Algorithms Peter & Clark skeleton algorithm (boot.skeleton()), based on pcalg and bnlearn R packages implementations. Peter & Clark (PC) algorithm (boot.pc()), based on pcalg and bnlearn R packages implementations. Fast Causal Inference (FCI), Really FCI (RFCI) and FCI+ algorithms (boot.fci()), based on pcalg R package implementation. Grow-Shrink (GS) algorithm (boot.gs()), based on bnlearn R package implementation. Incremental Association (IAMB), Fast IAMB, Interleaved IAMB and IAMB with FDR Correction algorithms (boot.iamb()), based on bnlearn R package implementation. Max-Min Parents & Children (MMPC), Semi-Interleaved Hiton-PC (SI-HITON-PC) and Hybrid Parents & Children (HPC) algorithms (boot.parents.children()), based on bnlearn R package implementation. Chow-Liu algorithm (boot.chowliu()), based on bnlearn R package implementation. Algorithm for the Reconstruction of Accurate Cellular Networks (ARACNE) (boot.aracne()), based on bnlearn R package implementation.

Score-Based Algorithms Hill-Climbing (HC) algorithm (boot.hc()), based on bnlearn R package implementation. Tabu Search (Tabu) algorithm (boot.tabu()), based on bnlearn R package implementation. Greedy Equivalence Search (GES) (boot.ges()), based on pcalg R package implementation. Linear NO-TEARS algorithm (boot.notears()), reimplemented in R.

Hybrid (Constraint-Based + Score-Based) Algorithms General 2-Phase Restricted Maximization (RSMAX2), Max-Min Hill Climbing (MMHC) and Hybrid HPC (H2PC) algorithms (boot.rsmax2()), based on bnlearn R package implementation. Adaptively Restricted Greedy Equivalence Search (ARGES) algorithm (boot.arges()), based on pcalg R package implementation.

Other Algorithms Graphical Lasso (GLASSO) (boot.glasso()), based on glasso R package implementation. Restricted Structural Equation Models (LINGAM) (boot.lingam()), based on pcalg R package implementation. GEne Network Inference with Ensemble of trees (GENIE3) (boot.genie3()), based on GENIE3 R package implementation. Graphical Continuous Lyapunov Models (GCLM), based on gclm R package implementation. * NODAG algorithm, based on Gherardo Varando implementation (https://github.com/gherardovarando/nodag).

Package Overview

Import & Export Files: import.dataset(): Import A Local Dataset import.geneset(): Import A Local Geneset import.graph(): Import A Local Graph export.dataset(): Export A Dataset To A File export.geneset(): Export A Geneset To A File export.graph(): Export A Graph To A File

RESTful API Client: list.datasets(): List Datasets Available Via RESTful API list.genesets(): List Genesets Available Via RESTful API list.graphs(): List Graphs Available Via RESTful API download.dataset(): Download A Dataset Via RESTful API download.geneset(): Download A Geneset Via RESTful API download.graph(): Download A Graph Via RESTful API

Dataset Manipulation: drop.all.zeros(): Drop rows and/or columns with all zeros filter.dataset(): Select (automatically) the most appropriate gene columns and cell rows of your dataset select.genes(): Select (automatically) the most appropriate gene columns of your dataset select.cells(): Select (automatically) the most appropriate cell rows of your dataset * dataset.split(): Split A Dataset Into Training And Test Datasets

Dataset Operations: gene.histogram(): Plot expression histograms of genes in your dataset gene.correlation(): Plot expression correlation of genes in your dataset * gene.clustering(): Gene Clustering

Structure Learning Algorithms: Without Bootstrapping: skeleton(): Peter & Clark Skeleton Algorithm pc(): Peter & Clark Algorithm (PC) fci(): Fast Causal Inference Algorithm (FCI) gs(): Grow-Shrink Algorithm (GS) iamb(): Incremental Association Algorithm (IAMB) parents.children(): Parents & Children Algorithm chowliu(): Chow-Liu Algorithm aracne(): ARACNE Algorithm hc(): Hill-Climbing Algorithm (HC) tabu(): Tabu Search Algorithm (TABU) ges(): Greedy Equivalence Search Algorithm (GES) notears(): Linear NO-TEARS Algorithm (Reimplemented) rsmax2(): General 2-Phase Restricted Maximization Algorithm (rsmax2) arges(): Adaptively Restricted Greedy Equivalence Search Algorithm (ARGES) glasso(): Graphical Lasso (GLASSO) lingam(): Restricted Structural Equation Models (LINGAM) genie3(): GEne Network Inference with Ensemble of trees (GENIE3) gclm(): Graphical Continuous Lyapunov Models (GCLM) * nodag(): NODAG Algorithm

With Bootstrapping: huge.graph(): Learn Huge Graph (With Random Gene Selection + Cells Bootstrapping) boot.skeleton(): Peter & Clark Skeleton Algorithm With Bootstrapping boot.pc(): Peter & Clark Algorithm (PC) With Bootstrapping boot.fci(): Fast Causal Inference Algorithm (FCI) With Bootstrapping boot.gs(): Grow-Shrink Algorithm (GS) With Bootstrapping boot.iamb(): Incremental Association Algorithm (IAMB) With Bootstrapping boot.parents.children(): Parents & Children Algorithm With Bootstrapping boot.chowliu(): Chow-Liu Algorithm With Bootstrapping boot.aracne(): ARACNE Algorithm With Bootstrapping boot.hc(): Hill-Climbing Algorithm (HC) With Bootstrapping boot.tabu(): Tabu Search Algorithm (TABU) With Bootstrapping boot.ges(): Greedy Equivalence Search Algorithm (GES) With Bootstrapping boot.notears(): Linear NO-TEARS Algorithm (Reimplemented) With Bootstrapping boot.rsmax2(): General 2-Phase Restricted Maximization Algorithm (rsmax2) With Bootstrapping boot.arges(): Adaptively Restricted Greedy Equivalence Search Algorithm (ARGES) With Bootstrapping boot.glasso(): Graphical Lasso (GLASSO) With Bootstrapping boot.lingam(): Restricted Structural Equation Models (LINGAM) With Bootstrapping boot.genie3(): GEne Network Inference with Ensemble of trees (GENIE3) With Bootstrapping boot.gclm(): Graphical Continuous Lyapunov Models (GCLM) With Bootstrapping boot.nodag(): NODAG Algorithm With Bootstrapping

Graph Format: detect.format(): Graph Format Detection convert.format(): Graph Format Conversion as.adjacency(): Convert Graph To Adjacency Matrix as.edges(): Convert Graph To Edge List as.graph(): Convert Graph To graph Format as.igraph(): Convert Graph To igraph Format * as.bnlearn(): Convert Graph To bnlearn Format

Graph Modifications: add.genes(): Add Genes To A Graph rename.genes(): Rename Genes Of A Graph delete.genes(): Delete Genes From A Graph delete.isolated(): Delete Isolated Nodes add.edges(): Add Edges To A Graph delete.edges(): Delete Edges To A Graph * fit.coefficients(): Estimate The Coefficients Of An Adjacency Matrix

Graph Operations: undirected.edges(): Return Undirected Edges directed.edges(): Return Directed Edges graph.plot(): Graph Plotting compare.graphs(): Graph Comparison gene.degree(): Degree Per Gene feature.degree(): Feature Degree Per Gene feature.plot(): Feature Graph Plotting graph.communities(): Graph Communities ortholog.graph(): Ortholog Genes Graph drugs.plot(): Drug-Gene Interactions Plotting average.graph(): Calculate The Average Graph score.graph(): Compute The Score Of A Graph

Graph generation: random.graph(): Generate A Random Graph/DAG ground.truth(): Create a Ground Truth Graph * make.edgelist(): Make an edgelist from some genes to anothers

Install gnlearn

if (!requireNamespace('devtools', quietly=TRUE))
    install.packages('devtools')
devtools::install_github('rlebron-bioinfo/gnlearn')

Docker Container

docker pull rlebronbioinfo/gnlearn:latest

Open an R interpreter with gnlearn already installed:

docker run -ti --rm -v "$PWD":/root/wd rlebronbioinfo/gnlearn R

Run an R script inside this Docker container:

cd path/to/script/directory
docker run -ti --rm -v "$PWD":/root/wd rlebronbioinfo/gnlearn Rscript /root/wd/myscript.R

After running one of these commands, your working directory will be mounted in /root/wd inside the Docker container. This container will be automatically destroyed after closing the interpreter.

Quick Start

Jupyter Notebook

Citation

The gnlearn article is under preparation. Please also use the following references according to the algorithm used:

• PC or skeleton [1]
• FCI [2]
• RFCI [3]
• FCI+ [4]
• GS [5]
• IAMB [6]
• Fast-IAMB [7]
• Interleaved-IAMB [6]
• IAMB with FDR Correction [8]
• MMPC [6]
• SI-HITON-PC [9]
• HPC [10]
• Chow-Liu algorithm [11]
• ARACNE [12]
• HC [13]
• Tabu [13]
• GES [14]
• Linear NO-TEARS [15] [16]
• RSMAX2 [17]
• MMHC [18]
• H2PC [10]
• ARGES [19]
• GLASSO [20]
• LINGAM [21]
• GENIE3 [22]
• GCLM [23]
• NODAG [24]

If you use a pcalg-based or bnlearn-based implementation, please cite [25] and [26] respectively.

References

[1] Colombo, D. and Maathuis, M. H. (2014). Order-independent constraint-based causal structure learning. The Journal of Machine Learning Research, 15(1):3741–3782.

[2] Spirtes, P., Glymour, C. N., Scheines, R., and Heckerman, D. (2000). Causation, Prediction, and Search. MIT press.

[3] Colombo, D., Maathuis, M. H., Kalisch, M., and Richardson, T. S. (2012). Learning high-dimensional directed acyclic graphs with latent and selection variables. The Annals of Statistics, pages 294–321.

[4] Claassen, T., Mooij, J., and Heskes, T. (2013). Learning sparse causal models is not np-hard. preprint arXiv:1309.6824.

[5] Margaritis, D. (2003). Learning bayesian network model structure from data. Technical report, Carnegie-Mellon Univ Pittsburgh Pa School of Computer Science.

[6] Tsamardinos, I., Aliferis, C. F., Statnikov, A. R., and Statnikov, E. (2003). Algorithms for large scale markov blanket discovery. In FLAIRS conference, volume 2, pages 376–380.

[7] Yaramakala, S. and Margaritis, D. (2005). Speculative markov blanket discovery for optimal feature selection. In Fifth IEEE International Conference on Data Mining, pages 1–4. IEEE.

[8] Peña, J. M. (2008). Learning gaussian graphical models of gene networks with false discovery rate control. In European conference on evolutionary computation, machine learning and data mining in bioinformatics, pages 165–176. Springer.

[9] Aliferis, F. C., Statnikov, A., Tsamardinos, I., Subramani, M. and Koutsoukos, X. D. (2010). Local Causal and Markov Blanket Induction for Causal Discovery and Feature Selection for Classification Part I: Algorithms and Empirical Evaluation. Journal of Machine Learning Research, 11:171–234.

[10] Gasse, M., Aussem, A., and Elghazel, H. (2014). A hybrid algorithm for bayesian network structure learning with application to multi-label learning. Expert Systems with Applications, 41(15):6755–6772.

[11] Chow, C. and Liu, C. (1968). Approximating discrete probability distributions with dependence trees. IEEE Transactions on Information Theory, 14(3):462–467.

[12] Margolin, A. A., Nemenman, I., Basso, K., Wiggins, C., Stolovitzky, G., Favera, R. D., and Califano, A. (2006). ARACNE: An Algorithm for the Reconstruction of Gene Regulatory Networks in a Mammalian Cellular Context. BMC Bioinformatics, 7(S1):S7.

[13] Bouckaert, R. R. (1995). Bayesian belief networks: from construction to inference. PhD thesis.

[14] Chickering, D. M. (2002). Learning equivalence classes of bayesian-network structures. Journal of machine learning research, 2(Feb):445–498.

[15] Zheng, X., Aragam, B., Ravikumar, P. K., and Xing, E. P. (2018). Dags with no tears: Continuous optimization for structure learning. In Advances in Neural Information Processing Systems, pages 9472–9483.

[16] Zheng, X., Dan, C., Aragam, B., Ravikumar, P., and Xing, E. P. (2019). Learning sparse nonparametric dags. preprint arXiv:1909.13189.

[17] Friedman, N., Nachman, I., and Pe’er, D. (2013). Learning bayesian network structure from massive datasets: The sparse candidate algorithm. preprint arXiv:1301.6696.

[18] Tsamardinos, I., Brown, L. E., and Aliferis, C. F. (2006). The max-min hill-climbing bayesian network structure learning algorithm. Machine learning, 65(1):31–78.

[19] Nandy, P., Hauser, A., and Maathuis, M. H. (2018). High-dimensional consistency in score-based and hybrid structure learning. The Annals of Statistics, 46(6A):3151–3183.

[20] Zhao, T., Liu, H., Roeder, K., Lafferty, J., and Wasserman, L. (2012). The huge package for high-dimensional undirected graph estimation in R. The Journal of Machine Learning Research, 13(1), 1059-1062.

[21] Shimizu, S., Hoyer, P. O., Hyvärinen, A., and Kerminen, A. (2006). A linear non-gaussian acyclic model for causal discovery. Journal of Machine Learning Research, 7(Oct):2003–2030.

[22] Huynh-Thu, V. A., Irrthum, A., Wehenkel, L., and Geurts, P. (2010). Inferring Regulatory Networks from Expression Data Using Tree-Based Methods. PLoS ONE, 5(9):e12776.

[23] Varando, G. and Hansen, N. R. (2020). Graphical continuous lyapunov models. arXiv preprint arXiv:2005.10483.

[24] Varando, G. (2020). Learning dags without imposing acyclicity. arXiv preprint arXiv:2006.03005.

[25] Kalisch, M., Mächler, M., Colombo, D., Maathuis, M. H., and Bühlmann, P. (2012). Causal inference using graphical models with the r package pcalg. Journal of Statistical Software, 47(11):1–26.

[26] Scutari, M. (2010). Learning Bayesian Networks with the bnlearn R Package. Journal of Statistical Software, 35(3).

rlebron-bioinfo/gnlearn documentation built on July 25, 2020, 12:38 p.m.