PMMdeNovo - Revision history

Eggeling at 09:49, 31 July 2017

2017-07-31T09:49:15Z

← Older revision		Revision as of 09:49, 31 July 2017
Line 1:		Line 1:
	__NOTOC__		__NOTOC__
	by Ralf Eggeling, Teemu Roos, Petri Myllymäki, and Ivo Grosse.		by Ralf Eggeling, Teemu Roos, Petri Myllymäki, and Ivo Grosse.

			== Paper ==
			The paper '''Inferring intra-motif dependencies of DNA binding sites from ChIP-seq data''' has been published in [http://www.biomedcentral.com/1471-2105/16/375 BMC Bioinformatics].

			'''Note:''' The software on this site is mainly intended to enable reproducibility of the results from the publication.
			For other purposes, please consider using the more recent [http://jstacs.de/index.php/InMoDe InMoDe] software, which contains the methodology of PMMdeNovo as well as more advanced features, speedups, better user interfaces, and automatic visualization.

	== Description ==		== Description ==

	=== Background ===		=== Background ===
	Statistical modeling of transcription factor binding sites is one of the classical fields in bioinformatics. The position weight matrix (PWM) model, which assumes statistical independence among all nucleotides in a binding site, used to be the standard model for this task for more than three decades but its simple assumptions are increasingly put into question. Recent high-throughput sequencing methods have provided data sets of sufficient size and quality for studying the benefits of more complex models. However, learning more complex models typically entails the danger of overfitting, and while model classes that dynamically adapt the model complexity to data have been developed, effective model selection is to date only possible for fully observable data, but not, e.g., within de novo motif discovery.		Statistical modeling of transcription factor binding sites is one of the classical fields in bioinformatics. The position weight matrix (PWM) model, which assumes statistical independence among all nucleotides in a binding site, used to be the standard model for this task for more than three decades but its simple assumptions are increasingly put into question. Recent high-throughput sequencing methods have provided data sets of sufficient size and quality for studying the benefits of more complex models. However, learning more complex models typically entails the danger of overfitting, and while model classes that dynamically adapt the model complexity to data have been developed, effective model selection is to date only possible for fully observable data, but not, e.g., within de novo motif discovery.
Line 9:		Line 17:
	=== Conclusions ===		=== Conclusions ===
	The traditional PWM model appears to be indeed insufficient to infer realistic sequence motifs, as it is on average outperformed by more complex models that take into account intra-motif dependencies. Moreover, using such models together with an appropriate model selection procedure does not lead to a significant performance loss in comparison with the PWM model for any of the studied transcription factors. Hence, we find it worthwhile to recommend that any modern motif discovery algorithm should attempt to take into account intra-motif dependencies.		The traditional PWM model appears to be indeed insufficient to infer realistic sequence motifs, as it is on average outperformed by more complex models that take into account intra-motif dependencies. Moreover, using such models together with an appropriate model selection procedure does not lead to a significant performance loss in comparison with the PWM model for any of the studied transcription factors. Hence, we find it worthwhile to recommend that any modern motif discovery algorithm should attempt to take into account intra-motif dependencies.

	~~== Paper ==~~
	~~The paper '''Inferring intra-motif dependencies of DNA binding sites from ChIP-seq data''' has been published in [http://www.biomedcentral.com/1471-2105/16/375 BMC Bioinformatics].~~

	== Runnable JARs ==		== Runnable JARs ==

Eggeling: links updated

2016-12-08T14:56:17Z

links updated

← Older revision		Revision as of 14:56, 8 December 2016
Line 18:		Line 18:

	=== ModelTrainer ===		=== ModelTrainer ===
	The tool [~~http~~://~~www2~~.~~informatik~~.~~uni-halle~~.de/~~agbio~~/~~publications~~/PMMdenovo/ModelTrainer.jar ModelTrainer] performs a de novo motif discovery on a set of putative non aligned sequences. It infers an inhomogenous PMM of arbitrary order, where order 0 corresponds to a PWM model.		The tool [https://www.cs.helsinki.fi/u/eggeling/PMMdenovo/ModelTrainer.jar ModelTrainer] performs a de novo motif discovery on a set of putative non aligned sequences. It infers an inhomogenous PMM of arbitrary order, where order 0 corresponds to a PWM model.
	Run by calling		Run by calling

Line 83:		Line 83:

	=== BindingSitePrediction ===		=== BindingSitePrediction ===
	The tool [~~http~~://~~www2~~.~~informatik~~.~~uni-halle~~.de/~~agbio~~/~~publications~~/PMMdenovo/BindingSitePrediction.jar BindingSitePrediction] predicts instances of binding sites in a positive data set based on a previously learned model.		The tool [https://www.cs.helsinki.fi/u/eggeling/PMMdenovo/BindingSitePrediction.jar BindingSitePrediction] predicts instances of binding sites in a positive data set based on a previously learned model.
	Run by calling		Run by calling

Line 131:		Line 131:

	=== Classification ===		=== Classification ===
	The tool [~~http~~://~~www2~~.~~informatik~~.~~uni-halle~~.de/~~agbio~~/~~publications~~/PMMdenovo/Classification.jar Classification] performs first a motif discovery with subsequent fragment-based classification by using positive data that is assumed to contain an instance of the motif, and negative data that is assumed not to contain the motif. This tool can be used for performing a single step of a K-fold cross validation experiment.		The tool [https://www.cs.helsinki.fi/u/eggeling/PMMdenovo/Classification.jar Classification] performs first a motif discovery with subsequent fragment-based classification by using positive data that is assumed to contain an instance of the motif, and negative data that is assumed not to contain the motif. This tool can be used for performing a single step of a K-fold cross validation experiment.
	Run by calling		Run by calling

Line 205:		Line 205:
	<td>The number of restarts of the algorithm.</td>		<td>The number of restarts of the algorithm.</td>
	</tr>		</tr>
	</table>		</table>
	The tool returns (i) the model complexity, i.e., the number of leaves of all parsimonious context trees of the learned motif model, and (ii) performance of the classifier measured by the area under the ROC curve.		The tool returns (i) the model complexity, i.e., the number of leaves of all parsimonious context trees of the learned motif model, and (ii) performance of the classifier measured by the area under the ROC curve.

Line 212:		Line 212:

	== Source code ==		== Source code ==
	Building the [~~http~~://~~www2~~.~~informatik~~.~~uni-halle~~.de/~~agbio~~/~~publications~~/PMMdenovo/PMMdenovo_sources.zip source code] requires Jstacs 2.1.		Building the [https://www.cs.helsinki.fi/u/eggeling/PMMdenovo/PMMdenovo_sources.zip source code] requires Jstacs 2.1.

Eggeling at 11:14, 10 November 2015

2015-11-10T11:14:18Z

← Older revision		Revision as of 11:14, 10 November 2015
Line 1:		Line 1:
	__NOTOC__		__NOTOC__
			by Ralf Eggeling, Teemu Roos, Petri Myllymäki, and Ivo Grosse.
			== Description ==
			=== Background ===
			Statistical modeling of transcription factor binding sites is one of the classical fields in bioinformatics. The position weight matrix (PWM) model, which assumes statistical independence among all nucleotides in a binding site, used to be the standard model for this task for more than three decades but its simple assumptions are increasingly put into question. Recent high-throughput sequencing methods have provided data sets of sufficient size and quality for studying the benefits of more complex models. However, learning more complex models typically entails the danger of overfitting, and while model classes that dynamically adapt the model complexity to data have been developed, effective model selection is to date only possible for fully observable data, but not, e.g., within de novo motif discovery.
			=== Results ===
			To address this issue, we propose a stochastic algorithm for performing robust model selection in a latent variable setting. This algorithm yields a solution without relying on hyperparameter-tuning via massive cross-validation or other computationally expensive resampling techniques. Using this algorithm for learning inhomogeneous parsimonious Markov models, we study the degree of putative higher-order intra-motif dependencies for transcription factor binding sites inferred via de novo motif discovery from ChIP-seq data. We find that intra-motif dependencies are prevalent and not limited to first-order dependencies among directly adjacent nucleotides, but that second-order models appear to be the significantly better choice.
			Conclusions
			=== Conclusions ===
			The traditional PWM model appears to be indeed insufficient to infer realistic sequence motifs, as it is on average outperformed by more complex models that take into account intra-motif dependencies. Moreover, using such models together with an appropriate model selection procedure does not lead to a significant performance loss in comparison with the PWM model for any of the studied transcription factors. Hence, we find it worthwhile to recommend that any modern motif discovery algorithm should attempt to take into account intra-motif dependencies.

	== Paper ==		== Paper ==
	The paper [http://www.biomedcentral.com/1471-2105/16/375 ~~Inferring intra-motif dependencies of DNA binding sites from ChIP-seq data] by Ralf Eggeling, Teemu Roos, Petri Myllymäki, and Ivo Grosse has been published in~~ BMC Bioinformatics.		The paper '''Inferring intra-motif dependencies of DNA binding sites from ChIP-seq data''' has been published in [http://www.biomedcentral.com/1471-2105/16/375 BMC Bioinformatics].

	== Runnable JARs ==		== Runnable JARs ==

Eggeling: paper published

2015-11-09T17:05:17Z

paper published

← Older revision		Revision as of 17:05, 9 November 2015
Line 1:		Line 1:
	__NOTOC__		__NOTOC__
	by Ralf Eggeling, Teemu Roos, Petri Myllymäki, and Ivo Grosse
			== Paper ==
			The paper [http://www.biomedcentral.com/1471-2105/16/375 Inferring intra-motif dependencies of DNA binding sites from ChIP-seq data] by Ralf Eggeling, Teemu Roos, Petri Myllymäki, and Ivo Grosse has been published in BMC Bioinformatics.

	== Runnable JARs ==		== Runnable JARs ==
Line 10:		Line 12:
	Run by calling		Run by calling

	<code>java -jar ~~InhPMM~~.jar inputFile motifWidth motifOrder flankingOrder initSteps addSteps restarts output</code>		<code>java -jar ModelTrainer.jar inputFile motifWidth motifOrder flankingOrder initSteps addSteps restarts output</code>

	where the arguments have the following semantics:		where the arguments have the following semantics:

Eggeling at 10:16, 23 February 2015

2015-02-23T10:16:11Z

← Older revision		Revision as of 10:16, 23 February 2015
Line 120:		Line 120:

	=== Classification ===		=== Classification ===
	The tool [http://www2.informatik.uni-halle.de/agbio/publications/PMMdenovo/Classification.jar Classification] performs first a motif discovery with subsequent fragment-based classification using positive data that is assumed to contain an instance of the motif, and negative data that is assumed not to contain the motif.		The tool [http://www2.informatik.uni-halle.de/agbio/publications/PMMdenovo/Classification.jar Classification] performs first a motif discovery with subsequent fragment-based classification by using positive data that is assumed to contain an instance of the motif, and negative data that is assumed not to contain the motif. This tool can be used for performing a single step of a K-fold cross validation experiment.
	Run by calling		Run by calling

Line 195:		Line 195:
	</tr>		</tr>
	</table>		</table>
	The tool returns the ~~classification results to~~ the ~~standard output~~.		The tool returns (i) the model complexity, i.e., the number of leaves of all parsimonious context trees of the learned motif model, and (ii) performance of the classifier measured by the area under the ROC curve.

	== Data ==		== Data ==

Eggeling at 13:28, 21 February 2015

2015-02-21T13:28:14Z

@@ Line 3: / Line 3: @@
 == Runnable JARs ==
-The application consists of three independent tools.
+The application consists of three independent tools. All tools have mandatory (no default values) and optional arguments.
 === ModelTrainer ===
@@ Line 119: / Line 120: @@
 === Classification ===
-The tool [http://www2.informatik.uni-halle.de/agbio/publications/PMMdenovo/Classification.jar Classification] performs first a motif discovery with subsequent fragment-based classification using positive data that is assumed to contain an instance of the motif, and negative data that is assumed not to contain the motif. The tool returns the classification results to the standard output.
+The tool [http://www2.informatik.uni-halle.de/agbio/publications/PMMdenovo/Classification.jar Classification] performs first a motif discovery with subsequent fragment-based classification using positive data that is assumed to contain an instance of the motif, and negative data that is assumed not to contain the motif.
 == Data ==

Eggeling at 13:05, 21 February 2015

2015-02-21T13:05:54Z

@@ Line 72: / Line 72: @@
 === BindingSitePrediction ===
 The tool [http://www2.informatik.uni-halle.de/agbio/publications/PMMdenovo/BindingSitePrediction.jar BindingSitePrediction] predicts instances of binding sites in a positive data set based on a previously learned model.
 === Classification ===

Eggeling at 11:59, 21 February 2015

2015-02-21T11:59:05Z

@@ Line 6: / Line 6: @@
 === ModelTrainer ===
-The tool [http://www2.informatik.uni-halle.de/agbio/publications/PMMdenovo/ModelTrainer.jar ModelTrainer] performs a de novo motif discovery on a set of putative non aligned sequences.
+The tool [http://www2.informatik.uni-halle.de/agbio/publications/PMMdenovo/ModelTrainer.jar ModelTrainer] performs a de novo motif discovery on a set of putative non aligned sequences. It infers an inhomogenous PMM of arbitrary order, where order 0 corresponds to a PWM model.
-It infers an inhomogenous PMM of arbitrary order, where order 0 corresponds to a PWM model.
+Run by calling
-The tool produces two output files, namely (i) an .xml file containing the learned model and (ii) a .dot file containing the graphViz representation of the learned PCT structures.
 === BindingSitePrediction ===

Eggeling at 11:36, 21 February 2015

2015-02-21T11:36:06Z

← Older revision		Revision as of 11:36, 21 February 2015
Line 1:		Line 1:
	=		__NOTOC__
			by Ralf Eggeling, Teemu Roos, Petri Myllymäki, and Ivo Grosse

			== Runnable JARs ==
			The application consists of three independent tools.

			=== ModelTrainer ===
			The tool [http://www2.informatik.uni-halle.de/agbio/publications/PMMdenovo/ModelTrainer.jar ModelTrainer] performs a de novo motif discovery on a set of putative non aligned sequences.
			It infers an inhomogenous PMM of arbitrary order, where order 0 corresponds to a PWM model.
			The tool produces two output files, namely (i) an .xml file containing the learned model and (ii) a .dot file containing the graphViz representation of the learned PCT structures.

			=== BindingSitePrediction ===
			The tool [http://www2.informatik.uni-halle.de/agbio/publications/PMMdenovo/BindingSitePrediction.jar BindingSitePrediction] predicts instances of binding sites in a positive data set based on a previously learned model.

			=== Classification ===
			The tool [http://www2.informatik.uni-halle.de/agbio/publications/PMMdenovo/Classification.jar Classification] performs first a motif discovery with subsequent fragment-based classification using positive data that is assumed to contain an instance of the motif, and negative data that is assumed not to contain the motif. The tool returns the classification results to the standard output.

			== Data ==
			The exemplary [http://www2.informatik.uni-halle.de/agbio/publications/PMMdenovo/data.tar.gz data sets] contain extracted ChIP seq sequences of 50 different human transcription factors from the [http://genome.ucsc.edu/ENCODE ENCODE project], as well as corresponding negative data. All data sets are split into 10 different subsets for enabling a reproducible 10-fold cross validation.

			== Source code ==
			Building the [http://www2.informatik.uni-halle.de/agbio/publications/PMMdenovo/PMMdenovo_sources.zip source code] requires Jstacs 2.1.

Grau: Created page with "="

2014-12-19T12:44:45Z

Created page with "="

New page