MD-SeeGH: a platform for integrative analysis of multi-dimensional genomic data

Bryan Chi, Ronald J. deLeeuw, Bradley P. Coe, Raymond T. Ng, Calum MacAulay, Wan L. LamBMC Bioinformatics 2008 · DOI

TL;DR

Analyzing the intricate interplay of genomic data (like copy number, methylation, and gene expression) from diverse sources is challenging. This paper introduces MD-SeeGH, a novel software platform designed to integrate and visualize these multi-dimensional datasets, enabling researchers to identify genetic alterations and compare complex genomic profiles across multiple experiments and groups.

What Problem Does This Paper Address?

Modern biological research generates vast amounts of genomic data from various high-throughput platforms. Understanding the downstream effects of genetic and epigenetic alterations on biological systems requires integrating data on DNA copy number status, DNA methylation, single nucleotide polymorphisms (SNPs), and gene expression. However, the heterogeneity of data types and platforms makes parallel analysis difficult. Existing tools often lack the comprehensive features needed for this integrative approach, creating a demand for robust software that can handle diverse genomic datasets and facilitate multi-dimensional analysis.

Methods

MD-SeeGH is built using Borland's C++ Builder6 and utilizes a MySQL backend database. It was developed and tested on Windows XP and Vista, with the software and documentation publicly available. The platform's core functionality revolves around:

  • Data Import and Preprocessing: MD-SeeGH accepts tab-delimited text files from various microarray platforms. It allows for dataset annotation, clinical information association, and seamless import of normalized data (e.g., from CGH-Norm).
  • Quality Assessment: The software includes features for filtering data based on standard deviation and signal-to-noise ratios. It also incorporates a tool to detect and compensate for recurrent artifacts in array CGH experiments.
  • Genetic Alteration Detection: MD-SeeGH supports the import of segmentation algorithm outputs (like CNA HMMer, DNAcopy, and aCGH Smooth) to identify regions of alteration within individual samples.
  • Gene and Annotation Tracks: Users can visualize various genomic annotations alongside experimental data, including RefSeq genes, microRNAs, CpG islands, and copy number variations. Information from external sources like OMIM and the UCSC Genome Browser can be displayed.
  • Multi-Sample Analysis: The platform allows for the alignment and direct comparison of up to 50 experiments, facilitating cross-platform analysis and visualization of multiple patient samples. It can also generate heatmaps for up to 100 samples to identify common regions of amplification or deletion.
  • Frequency Plots: MD-SeeGH can generate frequency plots for up to 1000 samples, highlighting minimal regions of interest across large datasets.
  • Group Comparison: The platform enables the comparison of two groups of experiments using frequency plots, with statistical tests (permutation testing, Fisher's exact tests) available for exported data.
  • Reporting and Export: Results can be exported as images (JPEG, bitmap), tab-delimited text files for further analysis, or formatted into ISCN reports for clinical use.

Datasets

The paper describes the use of various microarray platforms for evaluation, including:

  • RefSeq GeneTrack
  • SMRT array
  • Affymetrix GeneChip human mapping 500 K set
  • Agilent 244A
  • VUMC MACF human 30 K

The platform is designed to handle data from diverse experimental platforms, and the authors emphasize its ability to integrate data from both older and newer platforms.

Key Insights and Ideas

  • The Power of Integration: The central idea is that understanding complex biological systems requires looking at multiple genomic layers simultaneously. MD-SeeGH bridges the gap between disparate data types and platforms.
  • Flexibility is Key: The platform's design prioritizes user flexibility, allowing for easy updates of genome builds, import of various segmentation algorithms, and visualization of customizable annotation tracks.
  • Beyond Single-Sample Analysis: While single-sample analysis is crucial, the real power lies in comparing multiple samples and groups. MD-SeeGH's heatmap and frequency plot features are designed for this purpose.
  • Clinical Relevance: The ability to generate standardized ISCN reports directly from the software highlights its potential utility in clinical settings.
  • Addressing Artifacts: The inclusion of a tool to detect and compensate for recurrent artifacts in array CGH data is a significant practical improvement for data quality.

Key Results

The paper demonstrates MD-SeeGH's capability to:

  • Integrate and visualize data from multiple array platforms for the same sample (Figure 3).
  • Identify common regions of amplification or deletion across up to 100 samples using heatmaps (Figure 4).
  • Analyze and visualize alterations in up to 1000 samples using frequency plots (Figure 5).
  • Integrate different types of genomic data, such as copy number, SNP, and methylation profiles, for a single sample (Figure 6a,b).
  • Compare gene expression data with copy number alterations (Figure 6c).
  • Facilitate group comparisons of frequency plots to identify differences between disease types (Figure 7).
  • Generate ISCN reports for clinical applications (Figure 8).

The authors note that the Multiple Alignment feature can load 50 Agilent 244K arrays in approximately 1 minute and 20 seconds on a standard computer.

Limitations and Future Work

  • Statistical Analysis External: The paper explicitly states that statistical analysis for group comparisons must be done externally after exporting data from MD-SeeGH.
  • Non-Academic Use: There are restrictions on the use of MD-SeeGH by non-academics.
  • Future Development: While not explicitly detailed, the authors' emphasis on flexibility and the constant development of new analysis algorithms suggests ongoing improvements and potential for incorporating more advanced statistical methods directly into the platform. The ability to import new mapping information for future genome builds also points to a forward-looking design.

Summary updated

BibTeX
@article{chi2008md,
  author = {Bryan Chi and Ronald J. deLeeuw and Bradley P. Coe and Raymond T. Ng and Calum MacAulay and Wan L. Lam},
  title = {{MD-SeeGH: a platform for integrative analysis of multi-dimensional genomic data}},
  journal = {BMC Bioinformatics},
  year = {2008},
  doi = {10.1186/1471-2105-9-243},
}
Figure