Gseapy

Note: if you have limited RAM (1 core needs 1G), use v0.8.11.

* Fixed ``stack overflow`` bug in module ``ssgsea`` with v0.9.0.
- drop high dimensional numpy array for computing while using ``ssgsea``, but need more running time.
- ssgsea source code: revert back to old version of 0.8.11, which with multiprocessing support.

Notes about ssGSEA
-----------------------

* ES values from ``ssgsea`` are consistent with GSVA:gsva(method='ssgsea'), see [here](https://www.bioconductor.org/packages/release/bioc/html/GSVA.html).

* download **comparison.zip** and, run these two files to compare ``gseapy.ssgsea`` and GSVA:gsva(method='ssgsea')
- test.ssgesa.gseapy.py
- test.ssgesa.R.r


[comparison.zip](https://github.com/BioNinja/GSEApy/files/1592594/comparison.zip)

Next generation of GSEApy

``ssgsea`` using too much memory. Please use >=0.9.1

* Using high dimension numpy arrays to accelerate computation.
* drop multiprocessing, but reserved the keyword argument for future development.
* docs more clear
* APIs and internal structure of gseapy become more consistent and easier to maintain.
* minor bugs fixed

update some docs

* change matplotlib to export Type 2/TrueType fonts.
* add --no-scale option in command line
* add comments to the output raw.es, nes file
* ES values from ``ssgsea`` now are consistence with GSVA:gsva(method='ssgsea'), see [here](https://www.bioconductor.org/packages/release/bioc/html/GSVA.html)

bug fixed version

* improved the output file in ``ssgsea``:
- samples.raw.es.txt: enrichment score for all samples
- samples.normalized.es.txt: normalized enrichment score (nes) for all samples
- subfolders of each sample: nes, fdr, p-values are calculated by original GSEA method.
- **note**: enrichment plot only shows es

* ssgsea: scaled ES, add ``sample_norm_type`` argument to specify rank method. 36, rank method including:
- rank
- log_rank
- log
how these methods rank genes, see [ssGSEAprojection](http://rowley.mit.edu/caw_web/ssGSEAProjection/ssGSEAProjection.Library.R) , line 86.
* improved gct file input for ssgsea, gsea
- reset_index for pandas series index after sorting values (gct input)
* enrichr: fixed bug in single column input of dataframe
* docs improvement: new links, and more examples, see http://gseapy.rtfd.io/

Since 0.8.0, **multiprocessing** have been added to speed up calculation.
However, when multiprocessing mode is on, null distribution generated from random permutation are the same if each gene set has same gene member. This will affect a few users when input subsets only differ in enrich terms inside gene_set.gmt file. But from 0.8.5 and later, the broken null distribution are not issues any more.

**Users should use 0.8.5 and later to get most accurate results.**

For detail bugs:

* critical: fixed broken null distribution when permutation type is phenotype 28, thanks for iseekwonderful
* critical: now, fixed random permutation bug in Prerank and ssGSEA module. see 32.

And now, data input improvment
* ssGSEA:
- Accept a Series with gene names as index and return a dataframe 27
- Accept gene expression matrix in gct format 27
- Accept dataframe with only one column but gene names as index.
* Prerank:
- supports datafame with only one column but gene names as index
- supports Seires input
- dataframe with only two column, first col is gene name.
* Replot:
- fixed attribute no found bugs