Leveraging the new with the old: providing a framework for the integration of historic microarray studies with next generation sequencing

Abstract

Next Generation Sequencing (NGS) methods are rapidly providing remarkable advances in our ability to study the molecular profiles of human cancers. However, the scientific discovery offered by NGS also includes challenges concerning the interpretation of large and non-trivial experimental results. This task is potentially further complicated when a multitude of molecular profiling modalities are available, with the goal of a more integrative and comprehensive analysis of the cancer biology.

Figure 1

NGS Association System Information Architecture. The basis for integration of multiple molecular profiling modalities is illustrated. This includes processes for data transformation, reduction and association, as well as the direct interfacing to multiple custom and third party software tools and subsystems.

Figure 2

Samples and Molecular profiling studies. Illustrated is a diagram of the patient samples and molecular profiling studies used in the various analyses.

Figure 3

System Rendered Results for Key Cancer Biology Genes. Automatically rendered bar chart and table from a Cuffdiff dataset that was searched using two genes (KRAS and TP53), which have cancer biology significance in many cancers, including MM. The Cuffdiff Chart in subfigure A was quickly generated by a single point and click. All genes in the table (subfigure B) contain a significant p-value and q-value.

Figure 4

Triple Integration (WES, Cuffdiff Gene, Microarray). A triple integration was performed by combining Cuffdiff gene data with WES and a gene expression microarray experiment. The integrated WES data (green columns) includes: Effect, Codon Change, AA Change, Tumor DP (Depth), Tumor AF (Allelic Frequency). The corresponding RNA-seq data (light red columns), show the Sample 1 FPKM (pooled normal), Sample 2 FPKM (tumor), and p-value from Cufflinks. The microarray data (blue columns) display the corresponding Affy Probes, and Avg Probe Values. The grey columns are common across all modalities and show, Entrez ID, Gene Symbol, Gene Name, and Member List (gene list).

Figure 5

KRAS Automated Visualization and Exploration. Subfigure A shows an automated visualization of the FPKM Chart for KRAS and its associated isoforms. This was automatically generated by clicking the picture icon located in the rightmost region on the first line of the KRAS entry, in subfigure B. In this view, the novel isoform (i.e., CUFF.19733.1) was also fetched from the experimental data. The columns Transcript Id and P-Value show that only one transcript (ENST00000311936) is significant.

Figure 6

Validation of the KRAS G13D Hotspot Mutation via IGV Automation. IGV automation is utilized to validate the non-synonymous hotspot mutation involving codon 13 for KRAS, where Glycine is replaced by Aspartic Acid (G13D). By a single point and click on the web page entry for KRAS, the IGV display is rendered. During this process, all relevant information is spontaneously sent to IGV. This includes data from Variant Analysis, RNA-seq aligned reads, WES aligned reads, the reference genome and RABT assembly. For illustrative purposes, the significant isoform, ENST00000311936, is circled in green, and the location of the G13D hotspot mutation is circled in red. A potential novel isoform (CUFF.19733.1) from the normal pool is circled in blue.

Figure 7

KRAS Protein Plot. Automation is utilized to generate a "lolliplot", which illustrates the discovered amino acid changes for KRAS in the context of previously reported protein mutations and known domains. The G13D hotspot mutation found in the experiment is colored red. Other canonical mutations are colored blue. The RAS protein domain is colored green. Automation was achieved by a custom interface, along with R, the shiny package and Plot Protein.