epcy 0.0.1

Evaluattion of Predictive CapabilitY for ranking biomarker candidates.

Citing:

• EPCY: Evaluation of Predictive CapabilitY for ranking biomarker gene candidates. Poster at ISMB ECCB 2019: https://f1000research.com/posters/8-1349

Introduction:

This tool was developed to Evaluate Predictive CapabilitY of each feature to become a biomarker candidates.

Requirements:

• python3

• (Optional) virtualenv

Install:

python3 -m venv $HOME/.virtualenvs/epcy
source $HOME/.virtualenvs/epcy/bin/activate
cd [your_epcy_folder]
CFLAGS=-std=c99 pip3 install numpy==1.17.0
python3 setup.py install
epcy -h

Usage:

General:

From source:

cd [your_epcy_folder]
python3 -m epcy -h

After setup install:

epcy -h

Generic case:

• EPCY is design to work on any quantitative data, provided that values of each feature are comparable between each samples (normalized).

• To run a comparative analysis, epcy pred need two tabulated files:

  • A matrix of quantitative normalized data for each samples (column) with an “ID” column to identify each feature.

  • A design table which describe the comparison.

# Run epcy on any normalized quantification data
epcy pred -d ./data/small_for_test/design.tsv -m ./data/small_for_test/exp_matrix.tsv -o ./data/small_for_test/default_subgroup
# If your data require a log2 transforamtion, add --log
epcy pred --log -d ./data/small_for_test/design.tsv -m ./data/small_for_test/exp_matrix.tsv -o ./data/small_for_test/default_subgroup

• Result will be saved in prediction_capability.xls file, which is detail below.

• You can personalize the design file using –subgroup –query

epcy pred_rna -d ./data/small_for_test/design.tsv -m ./data/small_for_test/exp_matrix.tsv -o ./data/small_for_test/subgroup2 --subgroup subgroup2 --query A

Working on RNA sequencing readcounts:

• To run EPCY on readcounts not mormalized use pred_rna tool as follow:

# To run on read count not normalized, add --cpm --log
epcy pred_rna --cpm --log -d ./data/small_for_test/design.tsv -m ./data/small_for_test/exp_matrix.tsv -o ./data/small_for_test/default_subgroup

Working on kallisto quantification:

• EPCY allow to work directly on kallisto quantificaion using h5 files, to have access to bootstrapped samples. To do so, a kallisto column need to be add to the design file (to specify the directory path where to find abundant.h5 file for each sample) and epcy pred_rna need to run as follow:

# To run on kallisto quantification, add --kall (+ --cpm --log)
epcy pred_rna --kal --cpm --log -d ./data/small_leucegene/5_inv16_vs_5/design.tsv -o ./data/small_leucegene/5_inv16_vs_5/
# !!! Take care kallisto quantification is on transcript not on gene

• To run on gene level, a gff3 file of the genome annotation is needed, to have the correspondence between transcript and gene. This file can be download on ensembl

# To run on kallisto quantification and gene level, add --gene --anno [file.gff] (+ --kall --cpm --log)
epcy pred_rna --kal --cpm --log --gene --anno ./data/small_genome/Homo_sapiens.GRCh38.84.reduce.gff3 -d ./data/small_leucegene/5_inv16_vs_5/design.tsv -o ./data/small_leucegene/5_inv16_vs_5/

• kallisto quantification allow to work on TPM:

# work on TPM, replace --cpm by --tpm
epcy pred_rna --kal --tpm --log --gene --anno ./data/small_genome/Homo_sapiens.GRCh38.84.reduce.gff3 -d ./data/small_leucegene/5_inv16_vs_5/design.tsv -o ./data/small_leucegene/5_inv16_vs_5/

Output:

predictive_capability.xls

This file is the main output which contain the evaluation of each features (genes, proteins, …). It’s a tabulated files 9 columns:

• Default columns:

  • id: the id of each feature.

  • l2fc: log2 Fold change.

  • kernel_mcc: Matthews Correlation Coefficient (MCC) compute by a predictor using KDE.

  • kernel_mcc_low, kernel_mcc_high: boundaries of confidence interval (90%).

  • mean_query: mean(values) of samples specify as Query in design.tsv

  • mean_ref: mean(values) of samples specify as Ref in design.ts

  • bw_query: Estimate bandwidth used by KDE, to calculate the density of query samples

  • bw_ref: Estimate bandwidth used by KDE, to calculate the density of ref samples

• Using –normal:

  • normal_mcc: MCC compute a predictor using normal distributions.

• Using –auc –utest:

  • auc: Area Under the Curve

  • u_pv: pvalue compute by a MannWhitney rank test

• Using –ttest:

  • t_pv: pvalue compute by ttest_ind

subgroup_predicted.xls

Using –full a secondary output file (subgroup_predicted.xls) specify for each features if the sample as been correctly predicted. Build an heatmap with this output could help you to explore your data. More details coming soon.

Bagging:

To improve the stability and accuracy of MCC computed, you can add n bagging (using -b n)

#Take care, it's take n time more longer!!!, use multiprocess (-t) seems a good idea :).
epcy pred_rna -b 4 -t 4 --cpm --log -d ./data/small_for_test/design.tsv -m ./data/small_for_test/exp_matrix.tsv -o ./data/small_for_test/default_subgroup