Difference between revisions of "ABC-INT-Categorical features"
m (Created page with "<div id="ABC"> <div style="padding:5px; border:1px solid #000000; background-color:#e19fa7; font-size:300%; font-weight:400; color: #000000; width:100%;"> Integrator Unit: Cat...") |
m |
||
| Line 65: | Line 65: | ||
<!-- included from "./components/ABC-INT-Categorical_features.components.txt", section: "contents" --> | <!-- included from "./components/ABC-INT-Categorical_features.components.txt", section: "contents" --> | ||
| − | Most interesting data that describes function in living cells is not numerical, but categorical. | + | Most interesting data that describes function in living cells is not numerical, but categorical. Moreover, it is data with large numbers of categories - "'''high cardinality categorical data'''". Such data is problematic for machine learning for reasons of principle, and practicality. Such data is sparse, and the many dimensions of noise make overfitting of data a strong concern, in particular if we do not have very large numbers of examples in our training sets. The data suffers from the "curse of dimensionality", i.e. all examples look similarly similar or different. And the datastructures that hold such data may become impractically large, and model training may take impractically long. In this integrator unit we will download and prepare different types of categorical data to explore later how to use feature engineering to optimize it for machine learning tasks. |
Your tasks as a team are | Your tasks as a team are | ||
* to choose a dataset of interest for human systems biology; | * to choose a dataset of interest for human systems biology; | ||
| − | * to download the | + | * to download the source data; |
* to transform it into categorical features; | * to transform it into categorical features; | ||
* to submit your scripts and tools to the ''zu'' package; | * to submit your scripts and tools to the ''zu'' package; | ||
| Line 77: | Line 77: | ||
| − | + | === (1) Graph data mining on STRING=== | |
{{Smallvspace}} | {{Smallvspace}} | ||
| − | The [https://string-db.org/ '''STRING database'''] publishes a network of gene nodes and edges that represent functional interactions: these integrate various experimental observations and computational inference, such as protein-protein interactions and literature data mining - or they can be decomposed by individual | + | The [https://string-db.org/ '''STRING database'''] publishes a network of gene nodes and edges that represent functional interactions: these integrate various experimental observations and computational inference, such as protein-protein interactions and literature data mining - or they can be decomposed by individual categories of evidence. A summary score is given as a probability of an edge to be functionally relevant. Network data mining for ML features is a very interesting topic in and of itself, here we will simply take the neighbours of a gene as categorical features that describe its environment. Task: using a suitable score cutoff, produce a table of STRING neigbours for each human gene that is defined in our HUGO symbol table. Upload the annotation for the <tt>miniHUGOsymbols</tt> list to your documentation. |
| + | |||
| + | Example row: | ||
TNFRSF4 TNFRSF9|CTLA4|TNFSF4|TRAF5|IL2|IL2RA|FOXP3 | TNFRSF4 TNFRSF9|CTLA4|TNFSF4|TRAF5|IL2|IL2RA|FOXP3 | ||
| Line 86: | Line 88: | ||
{{Smallvspace}} | {{Smallvspace}} | ||
| − | + | === (2) GO and GOA=== | |
{{Smallvspace}} | {{Smallvspace}} | ||
| − | Gene Ontology Annotations provide the cornerstone of functional annotations for genes. Build a pipeline to annotate each HUGO symbol with the GO terms found in the relevant GOA tables. Task: produce a table of GO terms annotated for each human gene as defined by our HUGO symbol table. Do this separately for the three GO ontologies. Upload the annotation for the <tt>miniHUGOsymbols</tt> list to your documentation. Example header and row: | + | Gene Ontology Annotations provide the cornerstone of functional annotations for genes. Build a pipeline to annotate each HUGO symbol with the GO terms found in the relevant GOA tables. Task: produce a table of GO terms annotated for each human gene as defined by our HUGO symbol table. Do this separately for the three GO ontologies. Upload the annotation for the <tt>miniHUGOsymbols</tt> list to your documentation. |
| + | |||
| + | Example header and row: | ||
symbol MF BP CC | symbol MF BP CC | ||
| Line 95: | Line 99: | ||
{{Smallvspace}} | {{Smallvspace}} | ||
| − | + | === (3) MSigDB sets=== | |
{{Smallvspace}} | {{Smallvspace}} | ||
The Broad Institute hosts an expert-curated database of gene sets: [http://software.broadinstitute.org/gsea/msigdb/collections.jsp '''MSigDB''' - the Molecular Signature Database]. Task: download the data and build a pipeline to annotate all HUGO gene symbols with all of the gene sets that contain them. Annotate the <tt>miniHUGOsymbols</tt> list and upload that to your documentation and test how the pipeline scales to the full dataset of more than 17,000 gene sets. | The Broad Institute hosts an expert-curated database of gene sets: [http://software.broadinstitute.org/gsea/msigdb/collections.jsp '''MSigDB''' - the Molecular Signature Database]. Task: download the data and build a pipeline to annotate all HUGO gene symbols with all of the gene sets that contain them. Annotate the <tt>miniHUGOsymbols</tt> list and upload that to your documentation and test how the pipeline scales to the full dataset of more than 17,000 gene sets. | ||
| + | |||
Example row: | Example row: | ||
| Line 104: | Line 109: | ||
{{Smallvspace}} | {{Smallvspace}} | ||
| − | + | === (4) Enrichment=== | |
{{Smallvspace}} | {{Smallvspace}} | ||
A common aspect of systems biology wet-lab experiments is that they produce a ''set-of-genes'' result: genes that co-precipitate, genes that are co-regulated, genes that are phosporylated by the same kinase, etc. etc. Enrichment algorithms ask: what do such genes have in common, i.e. what feature appears more frequently in the set than one would expect in a randomly chosen set of genes. Any type of annotation can be chosen, but existing packages usually use GO annotations. Candidate tools include [http://bioconductor.org/packages/release/bioc/html/topGO.html topGO], and other tools in the [http://bioconductor.org/packages/release/BiocViews.html#___GeneSetEnrichment '''Gene Set Enrichment''' biocView]<ref>Note GSEA (Gene Set Enrichment Analysis) is '''not''' the same as gene feature enrichment.</ref>. Task: build a pipeline that takes as input a set of HUGO symbols - such as the sets derived from the MSigDB above, and outputs an annotation of enriched GO terms for each of the set elements. Develop this for the <tt>miniHUGOsymbols</tt> list and a few other gene sets and upload the results for the <tt>miniHUGOsymbols</tt> list to your documentation. | A common aspect of systems biology wet-lab experiments is that they produce a ''set-of-genes'' result: genes that co-precipitate, genes that are co-regulated, genes that are phosporylated by the same kinase, etc. etc. Enrichment algorithms ask: what do such genes have in common, i.e. what feature appears more frequently in the set than one would expect in a randomly chosen set of genes. Any type of annotation can be chosen, but existing packages usually use GO annotations. Candidate tools include [http://bioconductor.org/packages/release/bioc/html/topGO.html topGO], and other tools in the [http://bioconductor.org/packages/release/BiocViews.html#___GeneSetEnrichment '''Gene Set Enrichment''' biocView]<ref>Note GSEA (Gene Set Enrichment Analysis) is '''not''' the same as gene feature enrichment.</ref>. Task: build a pipeline that takes as input a set of HUGO symbols - such as the sets derived from the MSigDB above, and outputs an annotation of enriched GO terms for each of the set elements. Develop this for the <tt>miniHUGOsymbols</tt> list and a few other gene sets and upload the results for the <tt>miniHUGOsymbols</tt> list to your documentation. | ||
| + | |||
Example row: | Example row: | ||
| Line 113: | Line 119: | ||
{{Smallvspace}} | {{Smallvspace}} | ||
| − | + | === (5) Interpro=== | |
{{Smallvspace}} | {{Smallvspace}} | ||
[https://www.ebi.ac.uk/interpro/protein/P43489 '''InterPro''' provides] rich sequence and domain annotation - and the domain composition of a protein is a categorical feature set. [https://www.ebi.ac.uk/interpro/download.html '''Download'''] of InterPro data is available. Task: produce a table of InterPro domains in each human gene as defined by our HUGO symbol table. Upload the annotation for the <tt>miniHUGOsymbols</tt> list to your documentation. Example row: | [https://www.ebi.ac.uk/interpro/protein/P43489 '''InterPro''' provides] rich sequence and domain annotation - and the domain composition of a protein is a categorical feature set. [https://www.ebi.ac.uk/interpro/download.html '''Download'''] of InterPro data is available. Task: produce a table of InterPro domains in each human gene as defined by our HUGO symbol table. Upload the annotation for the <tt>miniHUGOsymbols</tt> list to your documentation. Example row: | ||
| Line 123: | Line 129: | ||
{{Vspace}} | {{Vspace}} | ||
| − | <div class="note">'''Do not upload your full datasets to the Github repository!'''</ | + | <div class="note">'''Do not upload your full datasets to the Github repository!'''</div> |
{{Vspace}} | {{Vspace}} | ||
Revision as of 02:48, 3 February 2018
Integrator Unit: Categorical Features
(Integrator unit: collect categorical features for human genes)
Abstract:
This page integrates material from the learning units and defines a task for defining and downloading categorical feature sets for human genes.
Deliverables:
- Integrator unit: Deliverables can be submitted for course marks. See below for details.
Prerequisites:
This unit builds on material covered in the following prerequisite units:
This unit is under development. There is some contents here but it is incomplete and/or may change significantly: links may lead to nowhere, the contents is likely going to be rearranged, and objectives, deliverables etc. may be incomplete or missing. Do not work with this material until it is updated to "live" status.
Contents
Evaluation
Your progress and outcomes of this "Integrator Unit" will be one of the topics of the first oral exam for BCB420/JTB2020. That oral exam will be worth 20% of your term grade.[1].
- Work through the tasks described below.
- Note that there are several tasks that need to be coordinated with your teammates and classmates. This is necessary to ensure the feature sets can be merged in the second phase of the course. Be sure to begin this coordination process in time.
- Remember to document your work in your journal concurrently with your progress. Journal entries that are uploaded in bulk at the end of your work will not be considered evidence of ongoing engagement. Note that this is a team task, and your contribution to the task must be clearly documented in your journal for evaluation.
- Your task will involve submitting documentation on a sub-page of the Teams and Tasks page. (Details below) This documentation will be jointly authored and I expect every team member to be able to speak to all of it.
- Your task will involve submitting code to the zu R package. Ensure that your team's submission are complete and pass package checks with zero errors, zero warnings and zero notes.
- Schedule an oral exam (if you haven't done so already) by editing the signup page on the Student Wiki. You must have signed-up for an exam slot before 20:00 on the day before your exam.[2]
- Your work must be complete before 20:00 on the day before your exam.
Contents
Most interesting data that describes function in living cells is not numerical, but categorical. Moreover, it is data with large numbers of categories - "high cardinality categorical data". Such data is problematic for machine learning for reasons of principle, and practicality. Such data is sparse, and the many dimensions of noise make overfitting of data a strong concern, in particular if we do not have very large numbers of examples in our training sets. The data suffers from the "curse of dimensionality", i.e. all examples look similarly similar or different. And the datastructures that hold such data may become impractically large, and model training may take impractically long. In this integrator unit we will download and prepare different types of categorical data to explore later how to use feature engineering to optimize it for machine learning tasks.
Your tasks as a team are
- to choose a dataset of interest for human systems biology;
- to download the source data;
- to transform it into categorical features;
- to submit your scripts and tools to the zu package;
- to document what you have achieved.
To begin, you need to choose - as a team - one of the following five data sources of categorical functional data:
(1) Graph data mining on STRING
The STRING database publishes a network of gene nodes and edges that represent functional interactions: these integrate various experimental observations and computational inference, such as protein-protein interactions and literature data mining - or they can be decomposed by individual categories of evidence. A summary score is given as a probability of an edge to be functionally relevant. Network data mining for ML features is a very interesting topic in and of itself, here we will simply take the neighbours of a gene as categorical features that describe its environment. Task: using a suitable score cutoff, produce a table of STRING neigbours for each human gene that is defined in our HUGO symbol table. Upload the annotation for the miniHUGOsymbols list to your documentation.
Example row:
TNFRSF4 TNFRSF9|CTLA4|TNFSF4|TRAF5|IL2|IL2RA|FOXP3
(2) GO and GOA
Gene Ontology Annotations provide the cornerstone of functional annotations for genes. Build a pipeline to annotate each HUGO symbol with the GO terms found in the relevant GOA tables. Task: produce a table of GO terms annotated for each human gene as defined by our HUGO symbol table. Do this separately for the three GO ontologies. Upload the annotation for the miniHUGOsymbols list to your documentation.
Example header and row:
symbol MF BP CC TNFRSF4 GO:0001618|GO:0005031|GO:0005515 GO:0006954|GO:006955|GO:0007275 GO:0005886|GO:0005887|GO:0009986
(3) MSigDB sets
The Broad Institute hosts an expert-curated database of gene sets: MSigDB - the Molecular Signature Database. Task: download the data and build a pipeline to annotate all HUGO gene symbols with all of the gene sets that contain them. Annotate the miniHUGOsymbols list and upload that to your documentation and test how the pipeline scales to the full dataset of more than 17,000 gene sets.
Example row:
TNFRSF4 M1739|M5947|M18255|M13664|M1644|M4248
(4) Enrichment
A common aspect of systems biology wet-lab experiments is that they produce a set-of-genes result: genes that co-precipitate, genes that are co-regulated, genes that are phosporylated by the same kinase, etc. etc. Enrichment algorithms ask: what do such genes have in common, i.e. what feature appears more frequently in the set than one would expect in a randomly chosen set of genes. Any type of annotation can be chosen, but existing packages usually use GO annotations. Candidate tools include topGO, and other tools in the Gene Set Enrichment biocView[3]. Task: build a pipeline that takes as input a set of HUGO symbols - such as the sets derived from the MSigDB above, and outputs an annotation of enriched GO terms for each of the set elements. Develop this for the miniHUGOsymbols list and a few other gene sets and upload the results for the miniHUGOsymbols list to your documentation.
Example row:
TNFRSF4 GO:0097190|GO:0051024|GO:0033209|GO:0032496
(5) Interpro
InterPro provides rich sequence and domain annotation - and the domain composition of a protein is a categorical feature set. Download of InterPro data is available. Task: produce a table of InterPro domains in each human gene as defined by our HUGO symbol table. Upload the annotation for the miniHUGOsymbols list to your documentation. Example row:
TNFRSF4 IPR034022|IPR001368|IPR001368
Process details TBC...
Notes
- ↑ Note: oral exams will focus on the content of Integrator Units, but will also cover material that leads up to it. All exams in this course are cumulative.
- ↑ For clarification: You sign up for only one oral exam for February.
- ↑ Note GSEA (Gene Set Enrichment Analysis) is not the same as gene feature enrichment.
Further reading, links and resources
- Taroni & Greene (2017) Cross-platform normalization enables machine learning model training on microarray and RNA-seq data simultaneously (BioRχiv doi: https://doi.org/10.1101/118349)
- Quantile Normalization is provided in the preprocessCore Bioconductor package:
Bolstad et al. (2003) A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics 19:185-93. (pmid: 12538238) [ PubMed ] [ DOI ] MOTIVATION: When running experiments that involve multiple high density oligonucleotide arrays, it is important to remove sources of variation between arrays of non-biological origin. Normalization is a process for reducing this variation. It is common to see non-linear relations between arrays and the standard normalization provided by Affymetrix does not perform well in these situations. RESULTS: We present three methods of performing normalization at the probe intensity level. These methods are called complete data methods because they make use of data from all arrays in an experiment to form the normalizing relation. These algorithms are compared to two methods that make use of a baseline array: a one number scaling based algorithm and a method that uses a non-linear normalizing relation by comparing the variability and bias of an expression measure. Two publicly available datasets are used to carry out the comparisons. The simplest and quickest complete data method is found to perform favorably. AVAILABILITY: Software implementing all three of the complete data normalization methods is available as part of the R package Affy, which is a part of the Bioconductor project http://www.bioconductor.org. SUPPLEMENTARY INFORMATION: Additional figures may be found at http://www.stat.berkeley.edu/~bolstad/normalize/index.html
- RNA-seq analysis is easy as 1-2-3 with limma, Glimma and edgeR Bioconductor workflow for RNAseq differential expression analysis with edgeR.
- RNA-seq workflow: gene-level exploratory analysis and differential expression Bioconductor workflow for RNAseq differential expression analysis with DEseq2.
- HUGO Gene Nomenclature Committee - the authoritative information source for gene symbols. Includes search functions for synonyms. aliases and other information, as well as downloadable data.
- Good discussion of current microarray normalization strategies, as well as a proposal how to apply QN to case/control datasets:
Cheng et al. (2016) CrossNorm: a novel normalization strategy for microarray data in cancers. Sci Rep 6:18898. (pmid: 26732145) [ PubMed ] [ DOI ] Normalization is essential to get rid of biases in microarray data for their accurate analysis. Existing normalization methods for microarray gene expression data commonly assume a similar global expression pattern among samples being studied. However, scenarios of global shifts in gene expressions are dominant in cancers, making the assumption invalid. To alleviate the problem, here we propose and develop a novel normalization strategy, Cross Normalization (CrossNorm), for microarray data with unbalanced transcript levels among samples. Conventional procedures, such as RMA and LOESS, arbitrarily flatten the difference between case and control groups leading to biased gene expression estimates. Noticeably, applying these methods under the strategy of CrossNorm, which makes use of the overall statistics of the original signals, the results showed significantly improved robustness and accuracy in estimating transcript level dynamics for a series of publicly available datasets, including titration experiment, simulated data, spike-in data and several real-life microarray datasets across various types of cancers. The results have important implications for the past and the future cancer studies based on microarray samples with non-negligible difference. Moreover, the strategy can also be applied to other sorts of high-throughput data as long as the experiments have global expression variations between conditions.
- Quackenbusch's paper is now old, but an often-cited standard reference in the field:
Quackenbush (2002) Microarray data normalization and transformation. Nat Genet 32 Suppl:496-501. (pmid: 12454644) [ PubMed ] [ DOI ] Underlying every microarray experiment is an experimental question that one would like to address. Finding a useful and satisfactory answer relies on careful experimental design and the use of a variety of data-mining tools to explore the relationships between genes or reveal patterns of expression. While other sections of this issue deal with these lofty issues, this review focuses on the much more mundane but indispensable tasks of 'normalizing' data from individual hybridizations to make meaningful comparisons of expression levels, and of 'transforming' them to select genes for further analysis and data mining.
If in doubt, ask! If anything about this learning unit is not clear to you, do not proceed blindly but ask for clarification. Post your question on the course mailing list: others are likely to have similar problems. Or send an email to your instructor.
About ...
Author:
- Boris Steipe <boris.steipe@utoronto.ca>
Created:
- 2018-02-01
Modified:
- 2018-02-01
Version:
- 0.1
Version history:
- 0.1 New unit under development
This copyrighted material is licensed under a Creative Commons Attribution 4.0 International License. Follow the link to learn more.
