MAgentOmics: Multi-agent Feature Selection for Integrative Multi-omics Analysis

This repository implements the MAgentOmics proposed in the paper “Multi-agent Feature Selection for Integrative Multi-omics Analysis” [1]. It is the first attempt to apply a multi-agent system for multi-view feature selection to handle the high-dimensionality nature of multi-omics data.

Reference:

[1] Sina Tabakhi and Haiping Lu, “Multi-agent Feature Selection for Integrative Multi-omics Analysis,” 2022 44th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Glasgow, Scotland, United Kingdom, 2022, pp. 1638-1642, doi: 10.1109/EMBC48229.2022.9871758.

Corrections

We have found one typo in the published paper. In Equation (1), we wrote the min operator instead of the max operator. Provided below is the correct format of the equation:

It should be noted that we have correctly implemented the equation in the source code.

Requirements

The code of the MAgentOmics method is implemented using Python 3.10.2. Below is the list of packages and frameworks we have used in the implementation:

• pandas 1.3.4
• numpy 1.20.3
• sklearn 0.24.2

Multi-omics Data

Ovarian serous cystadenocarcinoma data from the TCGA were selected to conduct the paper experiments that can be downloaded from the UCSC Xena Platform. Provided below are the characteristics of the Ovarian cancer multi-omics data.

Omics Type	Platform	Version	#Features	#Samples	Reference
DNA methylation	Illumina Infinium HumanMethylation27	2017-09-08	27,578	616	Link
Gene-level copy number variation	Affymetrix SNP 6	2017-09-08	24,776	579	Link
Gene expression RNA-seq	IlluminaHiSeq_RNASeqV2	2017-10-13	20,530	308	Link

When the data have been downloaded, you should change the following lines of code in the Jupyter Notebook to read data from your directory:

omics1 = pd.read_csv('DataSet_OvarianCancer/DNA_methylation',sep='\t',index_col=0)
omics2 = pd.read_csv('DataSet_OvarianCancer/genelevel_copy_number_alteration_CNA',sep='\t',index_col=0)
omics3 = pd.read_csv('DataSet_OvarianCancer/RNASeq',sep='\t',index_col=0)
label = pd.read_csv('DataSet_OvarianCancer/ClinicalMatrix',sep='\t',index_col=0)

Citation

If you find this method useful, please cite our paper:

@inproceedings{tabakhi2022multi,
  title={Multi-agent Feature Selection for Integrative Multi-omics Analysis},
  author={Tabakhi, Sina and Lu, Haiping},
  booktitle={2022 44th Annual International Conference of the IEEE Engineering in Medicine \& Biology Society (EMBC)},
  pages={1638--1642},
  year={2022},
  organization={IEEE}
  doi={10.1109/EMBC48229.2022.9871758}
}