Network mapping 101 course
0 likes42 views
The document outlines a course titled 'Network Mapping 101' that teaches how to integrate statistical and machine learning techniques for creating biochemical networks. Key topics include data preparation, statistical analysis, clustering, PCA, machine learning, and network analysis. The course includes various tutorials and resources to aid in understanding and applying these concepts.
Network mapping 101 course
- 1. Welcome to network mapping 101 Welcome to network mapping 101 In the following course you will learn how to integrate statistical, multivariate and machine learning results within a publication quality biochemical network. 1
- 2. Tutorials Tutorials https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/#topics 2 •Preparing raw data for analysis •Statistical analysis •Multivariate data exploration •Supervised clustering •Machine learning •classification •model validation •feature selection •Network analysis •biochemical •structural similarity •correlation •Network mapping – putting it all together
- 3. Analysis at the metabolomic scale 3
- 5. Identify what matters Univariate Group 1 Group 2 Multivariate Predictive Modeling ANOVA PCA PLS 5
- 6. Topics Topics https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/#topics 6 •Data preparation •Differential expression •Hierarchical Clustering •Principal Components Analysis (PCA) •Statistical analysis •Machine learning •Network analysis •Network mapping
- 7. How to think about data complexity n m 1-D 2-D m-D Data samples variables complexity Meta Data Experimental Design = Variable # = dimensionality 7
- 8. Data preprocessing Data preprocessing https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/partial/preprocess/ 8 define • data • row meta data • column meta data remove and/or impute • missing values
- 9. Follow along with the following tutorial: https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping _101/docs/partial/preprocess/ 9 Your turn Your turn
- 10. Differential expression Differential expression https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/partial/statistics/ 10 compare • class means identify • significant differences visualize • volcano plots • violin plots
- 11. Follow along with the following tutorial: https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping _101/docs/partial/statistics/ 11 Your turn Your turn
- 12. Hierarchical clustering (HCA) Hierarchical clustering (HCA) https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/partial/clustering/#heirarchical-clustering 12 group • samples and/or variables define similarity • correlation • distance • linkage visualize • heatmaps • dendrograms
- 13. Use the concept similarity/dissimilarity to group a collection of samples or variables approaches •hierarchical (linkage) •non-hierarchical (k-NN, k-means) •distribution (mixtures models) •density (DBSCAN) •self organizing maps (SOM) Linkage k-means Distribution Density 13 Clustering basics Clustering basics
- 14. identify •patterns •group structure •relationships evaluate and refine hypothesis reduce complexity Artist: Chuck Close 14 HCA goals HCA goals
- 15. distance • defines “nearness” or similarity X Y Euclidean X Y Manhattan Mahalanobis X Y * non-Euclidean HCA methods HCA methods
- 16. distance • defines “nearness” or similarity HCA methods HCA methods https://towardsdatascience.com/9-distance-measures-in-data-science-918109d069fa
- 17. HCA methods HCA methods single complete centroid average linkage or agglomeration • how samples or variables are connected or grouped
- 18. HCA methods HCA methods linkage or agglomeration https://scikit-learn.org/stable/auto_examples/cluster/plot_linkage_comparison.html
- 20. HCA interpretation HCA interpretation overview confirmation How does my metadata* match my data structure? * * *
- 21. Follow along with the following tutorial: https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping _101/docs/partial/clustering/#heirarchical-clustering 21 Your turn Your turn
- 22. https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/partial/multivariate/#pca 22 reduce • dimensionality maximize • variance explained visualize • variance explained • outliers • sample scores • variable loadings Principal Components Analysis (PCA) Principal Components Analysis (PCA)
- 23. James X. Li, 2009, VisuMap Tech. PCA goals PCA goals Principal Components (PCs) •non-supervised •projection of the data which maximize variance explained results 1. eigenvalues = variance explained 2. scores = new coordinates for samples (rows) 3. loadings = linear combination of original variables 23
- 24. PCA interpretation PCA interpretation Variance explained (eigenvalues) Row (sample) scores and column (variable) loadings 24
- 25. *no scaling or centering glucose 219021 25 PCA example PCA example
- 26. Relationship between scores and loadings Relationship between scores and loadings 26 loadings scores top loading variable’s scatterplot
- 27. Follow along with the following tutorial: https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping _101/docs/partial/multivariate/#pca Your turn Your turn 27
- 28. https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/partial/model/ 28 predict • sample classification optimize • model performance select • important features visualize • model performance • feature importance Machine learning Machine learning
- 29. 29 model validation Machine learning Machine learning https://scikit-learn.org/stable/modules/cross_validation.html
- 30. 30 model training and validation Machine learning Machine learning https://en.wikipedia.org/wiki/Learning_curve_(machine_learning)
- 31. https://www.researchgate.net/publication/359896672_A_Review_of_Machine_Learning_Methods_Applied_to_Structural_Dynamics_and_Vibroacoustic/figures?lo=1 31 bias vs. variance Machine learning Machine learning
- 32. https://en.wikipedia.org/wiki/Evaluation_of_binary_classifiers 32 classification performance Machine learning Machine learning
- 34. 34 feature selection Machine learning Machine learning https://scikit-learn.org/stable/modules/feature_selection.html#feature-selection
- 35. 35 Random Forest • cross-validation • feature selection Machine learning Machine learning https://mlu-explain.github.io/random-forest/
- 36. 36 Random Forest • decision path Machine learning Machine learning https://github.com/parrt/dtreeviz
- 37. 37 Random Forest • decision path Machine learning Machine learning https://github.com/parrt/dtreeviz
- 39. Follow along with the following tutorial: https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping _101/docs/partial/model/ 39 Your turn Your turn
- 40. https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/partial/network/ 40 network mapping • transform variables network calculation • regularized correlation • biochemical • structural similarity • model performance visualize • interactive networks Network analysis Network analysis
- 41. Components for network mapping Components for network mapping connections (edges) • empirical dependency (correlation) • biochemical (substrate/product) • chemical similarity • … nodes (vertices) • magnitude • importance • direction • relationships • … 41
- 42. Network data structures Network data structures adjacency matrix 42 https://www.steveclarkapps.com/graphs/
- 43. Network data structures Network data structures adjacency list 43 https://www.steveclarkapps.com/graphs/
- 44. Network data structures Network data structures edge list and vertices list (node attributes) 44 source target … index …
- 45. Correlation networks Correlation networks Connect molecules based on strength of their correlation or partial-correlation 45 bivariate multivariate network
- 46. regularized correlation network showing relationships in metabolic timeseries measurements for two classes of samples 46 Correlation example Correlation example
- 51. •Use structure to generate molecular fingerprint •Calculate similarities between metabolites based on fingerprint BMC Bioinformatics 2012, 13:99 doi:10.1186/1471-2105-13-99 Structural similarity networks Structural similarity networks 51
- 52. 52 Structural similarity example Structural similarity example
- 53. Treatment Effects Network 53 Metabolites Shape = increase/decrease Size = importance (loading) Color = correlation Connections violet = Biochemical relationships green = Structural similarity Combining networks Combining networks https://dgrapov.github.io/MetaMapR/gallery.html
- 54. Follow along with the following tutorial: https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping _101/docs/partial/network/ 54 Your turn Your turn
- 55. learn • Cytoscape basics map variables to • node attributes • edge attributes optimize • layout • legend • publication quality figure https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/partial/cytoscape/ 55 Network refinement and visualization Network refinement and visualization
- 56. Follow along with the following tutorial: https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping _101/docs/partial/cytoscape/ 56 Your turn Your turn