izMiR: computational ab initio microRNA detection
izMiR is an data analysis workflow that can be used for the detection of pre-miRNAs. The overall system includes two important workflows
1) Training of machine learning classifiers with suitable examples
2) Application of the learned model for detection of new pre-miRNAs.
By using the training workflow it is possible to generate models that can be used on new data to predict whether the given data have potential miRNA hairpins. It is also possible to use prediction workflow directly with the models and input data provided by us. In the latter case compatible features need to be calculated for analysis. These calculations can be done using two webservers. One provided by us (http://www.jlab.iyte.edu.tr/software/izmir) and one by Yones et al. (http://www.fich.unl.edu.ar/sinc/web-demo/mirnafe-full/).
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 10
Figure 11
Figure 12
In many machine learning based miRNA precursor prediction studies, different data sets were used with various features using different machine learning algorithms. Superficially comparing the published performance measures is at best misleading for the end users. To allow for a proper comparison, these tools need to be unified in one framework and tested on the exact same inputs. There are many challenges for such ab initio methods. Here we provide a comprehensive approach which allows the opportunity to compare different data sets, feature groups and classifiers. The system is very flexible and can be seamlessly adopted for future studies gracefully allowing extensions and adjustment of any settings.
We also showed that by using izMiR it is possible to obtain consensus models which lead to increased classification performance. For both learning and prediction features need to be calculated. For this two services are available one provided by Yones et al. (http://fich.unl.edu.ar/sinc/web-demo/mirnafe-full/) and one provided by us (http://jlab.iyte.edu.tr/software/izmir). The main parts of learning and prediction workflows are:
Training:
The below steps are laid out in the KNIME workflow (http://bioinformatics.iyte.edu.tr/software/izmir, training workflow) and the sections are labeled accordingly (Figures 1-6).
Prediction:
The below steps are laid out in the KNIME workflow (http://bioinformatics.iyte.edu.tr/software/izmir, prediction workflow) and the sections are labeled accordingly (Figures 7-12).
-
1) KNIME (https://www.knime.org) with all free extensions (at least the Weka plugin must be installed).
Training
1) KNIME installation
a. Download: https://www.knime.org/downloads/overview
b. Installation: https://tech.knime.org/installation-0
c. Update extensions: https://www.knime.org/downloads/update
2) Importing workflows
If you do not want to generate new models or results you can explore already computed results by right clicking on the nodes and choosing the output table for display.
If you want to make modifications to the workflow you can click on the nodes and change their settings. Some example changes could be:
Prediction
The prediction workflow requires a column named "Accession" for joining. If your data has no such columns you can use RowID node to create unique accession values.
Figures
Figure 1. Overall training workflow
Figure 2. MCCV and model generation
Figure 3. Sampling.
Figure 4. Studies (feature groups).
Figure 5. Feature selection and application of 3 classifiers.
Figure 6. Model sorting, selection and saving as PMML files.
Figure 7. Prediction workflow.
Figure 8 Prediction Meta-node
Figure 9 Decision Tree/Naïve Bayes Meta-node
Figure 10 Consensus Result Meta-node
Figure 11 Consensus Model Meta-node
Figure 12 Visualization Meta-node
-
-
-
-
-
Posted 15 Jul, 2016
Assigning scRNA-seq data to known and de novo cell types using CellAssign
Unified rational protein engineering with sequence-based deep representation learning
CUTseq is a versatile method for preparing multiplexed DNA sequencing libraries from low-input samples
Genome wide quantification of A-to-I RNA editing activity
Plating and imaging of yeast colonies for segmentation and classification with neural networks
Complex-centric proteome profiling by SEC-SWATH-MS for the parallel detection of hundreds of protein complexes
izMiR: computational ab initio microRNA detection
izMiR is an data analysis workflow that can be used for the detection of pre-miRNAs. The overall system includes two important workflows
1) Training of machine learning classifiers with suitable examples
2) Application of the learned model for detection of new pre-miRNAs.
By using the training workflow it is possible to generate models that can be used on new data to predict whether the given data have potential miRNA hairpins. It is also possible to use prediction workflow directly with the models and input data provided by us. In the latter case compatible features need to be calculated for analysis. These calculations can be done using two webservers. One provided by us (http://www.jlab.iyte.edu.tr/software/izmir) and one by Yones et al. (http://www.fich.unl.edu.ar/sinc/web-demo/mirnafe-full/).
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 10
Figure 11
Figure 12
In many machine learning based miRNA precursor prediction studies, different data sets were used with various features using different machine learning algorithms. Superficially comparing the published performance measures is at best misleading for the end users. To allow for a proper comparison, these tools need to be unified in one framework and tested on the exact same inputs. There are many challenges for such ab initio methods. Here we provide a comprehensive approach which allows the opportunity to compare different data sets, feature groups and classifiers. The system is very flexible and can be seamlessly adopted for future studies gracefully allowing extensions and adjustment of any settings.
We also showed that by using izMiR it is possible to obtain consensus models which lead to increased classification performance. For both learning and prediction features need to be calculated. For this two services are available one provided by Yones et al. (http://fich.unl.edu.ar/sinc/web-demo/mirnafe-full/) and one provided by us (http://jlab.iyte.edu.tr/software/izmir). The main parts of learning and prediction workflows are:
Training:
The below steps are laid out in the KNIME workflow (http://bioinformatics.iyte.edu.tr/software/izmir, training workflow) and the sections are labeled accordingly (Figures 1-6).
Prediction:
The below steps are laid out in the KNIME workflow (http://bioinformatics.iyte.edu.tr/software/izmir, prediction workflow) and the sections are labeled accordingly (Figures 7-12).
-
1) KNIME (https://www.knime.org) with all free extensions (at least the Weka plugin must be installed).
Training
1) KNIME installation
a. Download: https://www.knime.org/downloads/overview
b. Installation: https://tech.knime.org/installation-0
c. Update extensions: https://www.knime.org/downloads/update
2) Importing workflows
If you do not want to generate new models or results you can explore already computed results by right clicking on the nodes and choosing the output table for display.
If you want to make modifications to the workflow you can click on the nodes and change their settings. Some example changes could be:
Prediction
The prediction workflow requires a column named "Accession" for joining. If your data has no such columns you can use RowID node to create unique accession values.
Figures
Figure 1. Overall training workflow
Figure 2. MCCV and model generation
Figure 3. Sampling.
Figure 4. Studies (feature groups).
Figure 5. Feature selection and application of 3 classifiers.
Figure 6. Model sorting, selection and saving as PMML files.
Figure 7. Prediction workflow.
Figure 8 Prediction Meta-node
Figure 9 Decision Tree/Naïve Bayes Meta-node
Figure 10 Consensus Result Meta-node
Figure 11 Consensus Model Meta-node
Figure 12 Visualization Meta-node
-
-
-
-
-
Comments (4)
Comment moderated
on 25 November 2019
ab initio training online
commented on 26 November 2019
very nice information that you have shared
linux administration course online
commented on 26 November 2019
nice blog you have shared
ab initio training
commented on 27 November 2019
nice blog that you have shared