Rnalysis

-------------------
Version 4.1.0 of RNAlysis features improved performance and stability across the board, thanks to a switch to Qt6 and Numpy 2.


Changed
*******
* RNAlysis now runs on Qt6 instead of Qt5. This change should improve both performance and stability across RNAlysis.
* When importing parameters for differential expression analysis, RNAlysis will now automatically load the design matrix and chosen comparisons. This should work even with parameters that were exported in previous versions of RNAlysis.
* Made small improvements to the RNAlysis graphical interface.
* RNAlysis and its dependencies now run on Numpy 2 instead of Numpy 1.
* RNAlysis now uses a different implementation of the K-Medoids clustering algorithm, which should be more stable and faster than the previous implementation. However, note that the two implementations may give slightly different results.
* When running differential expression or feature counting, RNAlysis session reports will automatically include a logfile with R session info.
* Added optional parameters to all differential expression functions, allowing users to return a path to a logfile with R session info.

Fixed
******
* Fixed bug where table previews in the graphical interface would sometimes fail to update when applying functions inplace.

First, the software has switched from Pandas to Polars, which should substantially improve overall performance, particularly when generating automatic analysis reports.

Second, significant enhancements have been made to the automatic analysis reports.
These include new interactive features such as the ability to highlight analysis paths and filter the report by node types,
improved hierarchical layout for better readability, performance optimizations, and additional customization options.
The reports now offer a more comprehensive and user-friendly way to navigate and understand analysis workflows.

This update also includes various bug fixes and minor improvements to other functions within the software.

Happy analysis!

Added
******
* Added additional optional parameters to automatic report generation.
* Added additional parameters to 'Hierarchical clustergram plot' (CountFilter.clustergram).

Changed
*******
* RNAlysis now uses Polars instead of Pandas. This change should improve the performance of RNAlysis by an order of magnitude, especially when generating automatic analysis reports.
* Automatic analysis reports are now laid out in a hierarchical structure, making it easier to navigate through the report.
* When clicking on a node in an automatic analysis report, the analysis path leading to that node will be highlighted.
* When clicking on a legend node in an automatic analysis report, all nodes of that type will be highlighted.
* The 'Hierarchical clustergram plot' function should now run faster on large datasets.
* Added support for running tasks using 'sequential' backend on standalone releases of RNAlysis.

Fixed
******
* Fixed bug where the visualization functions 'Plot histogram of p-values' and 'Histogram of a column' would not display a graph immediately when using the stand-alone app.
* Fixed bug re-loading a saved session report would sometimes lead to missing connections in the analysis flow graph.
* Fixed bug where applying not-inplace operations to differential expression tables would change the expected name of the p-value column.
* Fixed bug where table previews in automatic analysis reports would display an extra empty line.

Backwards-incompatible changes
*******************************
* Since RNAlysis now uses Polars instead of Pandas, all functions that previously returned Pandas DataFrames will now return Polars DataFrames. To keep your code compatible with the new version, you may need to adjust the way you interact with the returned DataFrames, or convert them back into Pandas DataFrames.
* Sessions saved with previous versions of RNAlysis are mostly compatible with RNAlysis 4.0.0, but some of the new features of automatic analysis reports (such as highlighting the report by node type) may not work in older sessions.

-------------------
I'm happy to announce the latest release of RNAlysis, which brings a variety of new features and improvements.

One of the key additions in this version is expanded support for advanced differential expression analysis using DESeq2 and Limma-Voom.
You can now test continuous covariates, perform likelihood ratio tests for factors, interactions, and polynomials, and take advantage of sample quality weights in Limma-Voom analysis.
We've also added several new visualization options, such as a p-value histogram plot and more customization choices for gene expression plots.

This release also includes various usability enhancements and bug fixes to provide a smoother analysis experience. We've clarified error messages, improved parameter annotations, and resolved issues that could lead to crashes or incorrect results in certain scenarios. RNAlysis now integrates with the latest versions of the OrthoInspector and ShortStack APIs as well.

As always, your support and feedback are greatly appreciated.
If you have any questions, encounter issues, or have suggestions for future development, please don't hesitate to reach out.
Happy analysis!
Guy

Added
******
* Added support for advanced differential expression analysis with DESeq2/Limma-Voom, including testing continuous covariates, as well as likelihood ratio tests for factors, interactions, and polynomials.
* Added support for sample quality weights in Limma-Voom differential expression analysis.
* Added support for the 'cooksCutoff' parameter in DESq2 differential expression analysis.
* Added support for user-provided scaling factors in DESeq2 differential expression analysis.
* bowtie2 and kallisto paired-end modes now support the new file naming method 'smart', that attempts to automatically determine the common name of each pair of paired-end fastq files.
* Added a p-value histogram plot (DESeqFilter.pval_histogram) that displays the distribution of p-values in a differential expression table.
* Added new visualization options for 'plot expression of specific genes' function (CountFilter.plot_expression).
* Added new function 'Histogram of a column' (Filter.histogram) that generates a histogram of a column in a table.
* Added new function 'Concatenate two tables' (Filter.concatenate) that concatenates two tables along their rows.
* The filtering function 'Filter with a number filter' now supports filtering based on absolute values.

Changed
********
* When running differential expression analysis, RNAlysis will automatically ensure that the order of samples in your design matrix matches the order of samples in your count matrix, avoiding erronious results.
* The 'Filter genes with low expression in all columns' function (CountFilter.filter_low_reads) now supports the 'n_samples' parameter, allowing users to filter genes with a minimal expression threshold in a specific number of samples.
* The 'Plot expressino of specific genes' function (CountFilter.plot_expression) now supports the 'jitter', 'group_names' and 'log_scale' parameters, allowing users to further customize the plot.
* The 'Scatter plot - sample VS sample' function (CountFilter.scatter_sample_vs_sample) now always displays highlighted points on top of the plot, making it easier to see which points are highlighted.
* Kallisto quantification (kallisto_quantify_single_end and kallisto_quantify_paired_end) now supports the 'summation_method' parameter, allowing users to choose between 'raw' and 'scaled_tpm' transcript summation methods. The default behavior of the functions did not change (it corresponds to 'scaled_tpm').
* Enrichment bar plots now have optional parameters that control font sizes for titles and labels.
* Moved the enrichment analysis and enrichment graphs actions to the "Enrichment" menu in the graphical interface to make the actions easier to find.
* Improved the clarity of error messages when attempting to read an invalid GTF file.
* RNAlysis now supports the latest version of OrthoInspector API.
* RNAlysis now supports the latest version of Ensembl Ortholog API.
* Improved annotation for the 'metric' parameter of the 'Hierarchical clustergram plot' function (CountFilter.clustergram).
* Improved performance of RNAlysis when generating automatic session reports.
* RNAlysis now offers default values for differential expression tables' column names.
* Functions that average replicates now display clearer group names by default.
* The RNAlysis interface to ShortStack now uses the most recent API (replaced 'knownRNAs' with 'known_miRNAs').
* When running differential expression, RNAlysis session reports will automatically include the auto-generated R script, as well as the sanitized design matrix used.
* Added optional parameters to all differential expression functions, allowing users to return a path to the auto-generated R script and data sanitized design matrix used.

Fixed
*******
* Fixed bug where some FASTQ/SAM functions could not be added to a FASTQ pipeline.
* Fixed bug where bowtie2 could not be run in/from directories with spaces in their names.
* Fixed bug where RNAlysis would crash when launched without an internet connection.
* Fixed bug that cause ID-mapping functions to raise an error when called from the MacOS stand-alone app (thanks to `Mitchzw <https://github.com/Mitchzw>`_ in `#34 <https://github.com/GuyTeichman/RNAlysis/issues/34>`_).
* Fixed bug that caused R package installations (DESeq2, limma, etc) to fail on some computers (thanks to `Celine-075 <https://github.com/Celine-075>`_ in `#35 <https://github.com/GuyTeichman/RNAlysis/issues/35>`_).
* Fixed bug where Limma-Voom differential expression would fit numerical covariates as categorical factors.
* Fixed bug where generating enrichment bar plots with ylim='auto' would cause bars with 100% depletion (log2FC=-inf) to disappear.
* Fixed bug where defining 10 or more sample groups in the graphical interface would cause the groups to be ordered incorrectly in graphs.
* Fixed bug where the 'return_scaling_factors' argument would not return the normalization scaling factors on the graphical interface.
* Fixed various visual issues in the graphical interface
* Fixed bug where the 'filter_by_row_sum' function would raise an error when the table contains non-numerical columns
* Fixed bug where running enrichment on an empty gene set would raise an error.
* Fixed bug where RNAlysis would suggest resuming an auto-report from loaded session even when auto-report is turned off.
* Fixed bug where disabling auto-report in the middle of the session would raise errors when trying to create new graphs.
* Fixed bug where generating multiple gene expression plots (split_plots=True) with auto-generated report would only add the last graph to the session report.
* Fixed bug where the function 'normalize_to_quantile' generated unclear table names.
* Fixed bug where sometimes the first operation performed in a session would not display correctly in the automatic session report.

New Contributors
*****************
* `Mitchzw`_ in `34`_
* `Celine-075`_ in `35`_

-------------------
This release brings several exciting new features.
Notably, these inclue the ability to run functions from the Picardtools suite, and the ability to automatically save interactive session reports and later resume them from any saved session.
In addition, this release includes several visual upgrades, bug fixes, and quality-of-life improvements.
Happy analysis!

Added
*******
* *RNAlysis* can now run functions from the Picardtools suite, including conversion functions (BAM to SAM, SAM to FASTQ, FASTQ to SAM, etc), quality control (validate SAM), and post-processing functions (remove PCR duplicates, sort SAM, create BAM index).
* *RNAlysis* interactive session reports can now be resumed from any saved session, instead of having to start a new report from scratch. When loading a session created by *RNAlysis* 3.11 and beyond, you will have the option to resume the interactive report from the last saved state.

Changed
********
* CutAdapt adapter trimming functions can now receive an optional "new_filenames" parameter, which allows users to specify the names of the output files.
* Hierarchical clustergram plot (CountFilter.clustergram) now supports the 'colormap' parameter, which allows users to specify a custom color map for the plot.
* Hierarchical clustergram plot (CountFilter.clustergram) now displays continuous values on the color bar, instead of discrete values.
* Generally improved the looks of Hierarchical clustergram plot (CountFilter.clustergram).
* Previously-added functions (such as ortholog/paralog mapping, biotype summary by GTF file, etc) can now be applied to gene sets. Previously, some of these functions could only be applied to data tables.

Fixed
*******
* Function "Summarize feature biotypes (based on a reference table)" (biotypes_from_ref_table) now treats rows with missing values as "_missing_from_biotype_reference" instead of ignoring them entirely.
* Fixed bug where the Ensembl paralog-finding function would appear under the wrong tab in the graphical interface.
* Fixed bug where the description of the MA Plot function and parameters would not display correctly in the graphical interface.

-------------------
Version 3.10.1 introduces several bug fixes, as well as well as support for random effect analysis in Limma-Voom differential expression.

Added
*******
* Limma-Voom differential expression can now fit mixed linear models containing a random effect (e.g. nested design).


Fixed
*******
* Fixed bug where trying to load sessions created with RNAlysis version 3.10.0 would result in an error.
* Fixed bug where using the OrthoInspector ortholog mapping function with database='auto' would sometimes fail to find an appropriate mapping database, even when one exists.
* Fixed bug where kallisto paired-end quantification window would display the 'read1' and 'read2' parameters twice.
* Fixed bug where empty sub-menus would appear under the FASTQ menu. These sub-menus will be implemented in future versions of RNAlysis.

-------------------
I'm thrilled to introduce RNAlysis version 3.10.0.
This version includes features that were requested by users for a while, alongside quality-of-life improvements and bug fixes.
Here is a brief highlight of the most important additions:

**Ortholog Mapping:** *RNAlysis* can now map genes to their closest orthologs in different organisms.
You can map genes to their orthologs using four different databases - Ensembl, Panther, PhylomeDB, and OrthoInspector - extracting both one-to-one and one-to-many ortholog relationships and filtering them based on their reliability.

**Discovering Paralogs:** In the same vein, *RNAlysis* now facilitates the discovery of paralogs within a specific organism, using either the Ensembl or Panther databases.

**New visualization and analysis options for Principal Component Analysis (PCA):** We've introduced new functions and parameters to allow you to get more out of your principal component analysis.

I would also like to extend my personal apology for the delay in bringing you this update.
Due to personal reasons, this release, originally scheduled for the end of August, took longer than expected.
Your patience and support have been invaluable, and I'm eager to share these exciting additions with you.
Thank you for being a part of the RNAlysis community, and stay tuned for more updates in the near future!

Added
*******
* Added new functions to the filtering module that map genes to their closest orthologs in a different organism, using four different databases: Ensembl, Panther, PhylomeDB, and OrthoInspector.
* Added new functions to the filtering module that find paralogs of genes in a given organism, using two different databases: Ensembl and Panther.
* Added new function 'Sort table by contribution to a Principal Component (PCA)' (CountFilter.sort_by_principal_component), which allows sorting of genes in a count matrix by their contribution (gene loadings) to a principal component.
* Added a new parameter called 'legend' to 'Principal Component Analysis (PCA) plot' (CountFilter.pca), which allows users to display a legend on the PCA plot with a name for each sample group/color.

Changed
********
* RNAlysis now issues a warning when users run PCA or PCA-based functions on an unnormalized count matrix.
* The 'seek_fusion_genes' and 'learn_bias' arguments for kallisto quantification (fastq.kallisto_quant_single_end and fastq.kallisto_quant_paired_end), which were depracated in kallisto versions >0.48, are no longer displayed on the graphical interface. Old Pipelines that contain these arguments will still run, but new Pipelines will not contain them.
* Long-running functions now run in the background even when the 'inplace' parameter is set to True, instead of freezing the entire graphical interface.

Fixed
*******
* Fixed bug where functions would sometimes fail to run without displaying an error message.
* Fixed bug where progress bars in the graphical interface would sometimes not disappear after reaching 100% completion.
* RNAlysis should no longer display warning messages about graph layout when graphs are scaled down.
* Fixed bug where the clustergram function (CountFilter.clustergram) would raise an error with specific sets of dependecy versions.
* Loading tables no longer raises a depracation warning when using newer versions of Pandas.