comPyDEtools

A Python implementation of a part of compareDEtools.

comPyDEtools can ...

1. Generate simulated dataset (KIRC, Bottomly, mKdB or mBdK)
2. Run DE analysis (using subprocess.run())
3. Generate the figures like Fig 2 in Baik 2020

and can't ...

• SEQC benchmark (like Fig 1 in Baik 2020)
• False positive count comparison (like Fig 3 in Baik 2020)
• etc

Installation

pip install https://github.com/136s/comPyDEtools.git

Usage

1. Make a condition file like compydetools/data/synthetic_conditions.yaml.

2. Run

   python -m compydetools condition.yaml  # specify your condition file made at step 1.

   or

   run in Python

   from compydetools.condition import CONDITION, set_condition
   from compydetools.core import Paper
   from compydetools.utils import run_commands
   
   set_condition("condition.yaml")  # specify your condition file made at Usage 1.
   paper = Paper(nrep=CONDITION.nrep)
   paper.generate_datasets()
   for anal_res in run_commands(CONDITION.analysis.cmds):
       print(anal_res)
   paper.make()

3. Check generated files

   • input/: simulated RNA-seq data
     • dataset structure
       • first line is header
       • Gene_ID column: sequential numbers from 1 to the number of genes
       • Gene_Symbol column: "LOC" + Gene_ID
       • Description column: "up" (upregulated), "dn" (downregulated) or "ns" (not significant)
       • remaining columns: simulated expression counts for each samples and smaple names are "TRT-*" (treatment sample) or "CTRL-*" (control sample) (* is a sequential number for each condition)
     • dataset property
       • file path: {simul_data}_{disp_type}_upFrac{frac_up}_{nsample}spc_{outlier_mode}_{nde}DE/{simul_data}_{disp_type}_upFrac{frac_up}_{nsample}spc_{outlier_mode}_{nde}DE_rep{seed}.tsv
       • newline character: LF
       • enxoding: UTF-8
   • result/: plots of performance comparison

condition file (condition.yaml)

• analysis: configuration of DE analysis
  • cmds: a list of DE analysis commands
  • res: a reguler expression of a path to result files
    • "{count_stem}" replaced by dataset path stem
    • "{method_type}" replaced by method_type
  • de_true: column name of deg regulation (up, dn or ns) in each result files (defaults to "Description")
  • de_score: column name of deg score like p-value in each result files (defaults to "padj")
  • de_score_threshold: threshold of de_score (DEGs' de_score is lower than de_score_threshold)
• dirs: directories of generated files
  • dataset: generated simulated datasets
  • result: plots of performance comparison, csv of metrics values and pickle of Paper instance
• simul_data: KIRC, Bottomly, mKdB or mBdK
• disp_type: same or differnt
• frac_up: fraction upregulated in DEGs (float, $[0, 1]$)
• nsample: number of samples per groups (int, 3<=)
• outlier_mode: D, R, OS, or DL
• pde: percent of DE in all genes (float, $(0, 100]$)
• metrics_type: auc, tpr, fdr, cutoff, f1score or kapppa
  • if you want to add any metrics, modify const.Metrics and utils.calc_metrics() by fork or PR
• method_type: specify your DE analysis method (defaults to {"deseq2": "Deseq2"})
  • comPyDEtools recognizes the type of DE analysis method only by the output folder path (analysis.res in the condition file)
• nrep: number of simulation repetition under one condition (int, $3&lt;=$)

Development

Class relationships

erDiagram
  Paper |o--|{ Figure : "has a list of"
  Figure |o--|{ Plot : "has a list of"
  Plot ||--|{ DataPool : "has a list of"
  DataPool ||--|{ Dataset : "has a list of"
  DataPool ||--|{ Result : "has a list of"
  Dataset ||--|| Result : ""

  Paper {
    int nrep "number of repetition in a data pool (3<=)"
    int seed "global random seed"
    list[Figure] figures
  }
  Figure {
    Simul simul_data PK "simulation data (KIRC, Bottomly, mKdB or mBdK)"
    Disp disp_type PK "dispersion type (same or differnt)"
    float frac_up PK "fraction upregulated ([0, 1])"
    list[Plot] plots
  }
  Plot {
    int nsample PK "number of samples per condition (3<=)"
    Outlier outlier_mode PK "outlier mode (D, R, OS, or DL)"
    list[DataPool] datapools
  }
  DataPool {
    float pde PK "percent of DE in all genes ((0, 100])"
    list[Dataset] datasets
    list[DataPool] datapools
  }
  Dataset {
    int seed PK "random seed for each dataset generated from global seed"
    DataFrame counts "simulated count matrix"
  }
  Result {
    int seed PK "random seed for each dataset"
    list[Method] method_types "a list of DE analysis methods to be compared"
    list[Metrics] metrics_types "a list of metrics to comprere DE analysis methods"
  }

Loading

core module

• Paper class represents all figures in the condition file
• Figure class represents a figure (like Fig 2)
• Plot class represents a sub figure (like Fig 2A)
• DataPool class represents same condtion datasets (contains nrep datasets)
• Dataset class represents a simulated count matrix
• Result class represents a results of a Dataset under each method and metrics

property | Class	Paper	Figure	Plot	DataPool	Dataset	Result
a list of	Figure	Plot	DataPool	Dataset, Result
number of repetition (nrep)	1	1	1	1
simulation data (simul_data)		1	1	1	1	1
dispersion type (disp_type)		1	1	1	1	1
fraction upregulated (frac_up)		1	1	1	1	1
number of samples (nsample)			1	1	1	1
outlier mode (outlier_mode)			1	1	1	1
percent of DE in all genes (pde)				1	1	1
simulated count matrix					1	1
method type (method_type)						*
metrics type (metrics_type)						*

Table: Class / property correspondence (*: many)

const module

• Simul class is a list of simulation dataset names
  • simul_data in the condition file
• Disp class is a list of dispersion condition
  • disp_type in the condition file
• Outlier class is a list of outlier mode
  • outlier_mode in the condition file
• Metrics class is a list of metrics of performance comparison
  • metrics_type in the condition file
• Method class is a list of DE analysis method
  • method_type in the condition file

License

GPL2.0

This is a partial port of unistbig/compareDEtools.