Chapter 3: Use Case with Lasergene Genomics & Mastermind Genomic Search Engine

A recent collaboration between DNASTAR and Resonant Therapeutics used RNA-Seq Illumina reads to study gene expression differences in ovarian carcinoma cells (OVCAR-2 cell line) grown using 2D and 3D cultures. In case these terms are unfamiliar, a 2D culture is where cells are grown in a petri dish, while a 3D culture is grown on a matrix that mimics the microenvironment of a tumor. In this study, we show how to quickly reduce a large data set down to three variants of interest, and then demonstrate the utility of Genomenon’s Mastermind Genomic Search Engine for exploring deleterious variants.

Setting up and running the assembly in SeqMan NGen

SeqMan NGen was used to align Illumina RNA-Seq data from multiple replicates to the GRCh38. p2 human genome package. In addition to read alignment, the following steps were performed automatically during the assembly process:

1. SNPs and other small variants were called using the diploid model
2. Variants were annotated using the Variant Annotation Database (VAD)
3. Differential gene expression values were calculated using DESeq2

This study involved cancer cells, which understandably have a large number of mutations. During assembly, SeqMan NGen called 965,209 variants.

Filtering Variants in ArrayStar

After assembly in SeqMan NGen, we used the Advanced Filtering feature in ArrayStar to identify variants found in all six experiments that met the following criteria (Figure 6, next page):

• General tab – Non-synonymous SNPs only
• Statistics tab – Pnotref > 90%, SNP% > 15, Depth > 20
• Pathogenicity tab – Labeled in the ClinVar database as Pathogenic or Likely Pathogenic

Figure 6. Selections made in ArrayStar’s “Advanced Filtering” dialog.

Figure 6a. Filtering dialog.

Figure 6b. General filtering options with nonsynonymous filter applied.

Figure 6c. Statistical SNP filtering options.

Figure 6d. Pathogenecity filtering options with ClinVar filter applied.

ArrayStar found three variants that fit all criteria (Figure 7).

Figure 7. The three variants of interest found after filtering in ArrayStar.

Looking at these results in the Variant Table, we were able to see the SNP base and reference bases, the number of articles found in Mastermind, links to that evidence, the clinvar_trait, and the MutationTaster_pred. For legibility in this ebook, the Variant Table has been split in half to create two figures (Figures 8 and 9).

Figure 8. The leftmost columns in ArrayStar’s Variant Table.

Figure 9. The rightmost columns in ArrayStar’s Variant Table, along with a tooltip for the selected variant and a view of the Details panel.

Exploring Genomic Associations with Mastermind

Note that the variant in the KRAS gene (middle of the three variants) has over 4,300 Mastermind citations. This well-known variant, located on chromosome 12 at position 25245350, causes the G12V change in the KRAS protein and is the key driver mutation behind the tumors being studied. In addition to the two-star status of this SNP in ClinVar, notice that all three functional prediction methods present in the VAD also flag it as deleterious. As shown near the bottom of the tooltip, this KRAS gene mutation has been associated with a wide variety of cancers.

ArrayStar includes a direct link to Mastermind, where we can view the entry for known KRAS mutations (Figure 10).

Figure 10. The Mastermind Professional Edition landing page for the KRAS G12V variant.

Optimize output with prioritized sorting and filtering

The image on the previous page shows the full list of over 4,300 publications citing this variant. The literature base can then be focused by adding phenotype keywords. In this case, when “ovarian carcinoma” is applied, the reference number is reduced to 140 articles (Figure 11).

Figure 11. Mastermind KRAS G12V with ovarian carcinoma phenotype applied.

Other filters may be used to further refine this search, including expression filters in the “Genetic
Mechanism” tab (Figure 12).

Figure 12. Mastermind Filters including expression.

This focuses the results on 65 highly relevant articles (Figure 13).

Figure 13: Mastermind KRAS G12V Filtered by phenotype, expression and relevance.

The articles displayed are sorted by relevance, with the larger circles in the publication history pane denoting the most relevant articles. Relevance is determined by several factors, including, but not limited to:

• The number of times in which the gene (“KRAS”) and variant (“G12V”) names are mentioned in
  each article,
• Whether these terms are in the title, abstract and/or full text or supplemental information, and
• Proximity to phenotype and other keywords applied through the filtering process.

All of these elements can be viewed in the “Full Text Matches” section by viewing all matches being
highlighted in the sentence fragments.

Uncover data connections with genomic associations

Due to the highly complex nature of genetic disorders and the variety of factors that contribute to their pathogenicity, the opportunity to create clarity around how your search criteria of interest relate with each other is exceedingly beneficial. The Genomic Associations functionality within Mastermind allows you to explore significant genetic associations that may otherwise be missed or unrecognized, giving you the ability to uncover key data that may lead to a new hypothesis, a research discovery, or even solving a clinical case.

To expand on the example, suppose that you are interested in exploring treatment options for ovarian carcinoma related to KRAS p.G12V mutations. Clicking on the blue “Explore Associations” box (Figure 14) directs you to a flexible interface that connects your search keywords with all associated genetic factors in the medical evidence.

Figure 14. The Explore Associations box on the Mastermind page.

In this case, the “Therapies” tab reveals prioritized articles covering therapeutic intervention for KRAS p.G12V-related ovarian carcinoma. As seen in Figure 15, in terms of article quantity, the top three pharmacotherapies are:

1. Paclitaxel
2. Selumetinib
3. Carboplatin

Figure 15. The “Therapies” tab with the top three pharmacotherapies boxed in orange.

Awareness of therapies surrounding a particular genetic variant creates informed focus during hypothesis formation and study design. Most importantly, however, this invaluable insight significantly impacts clinical decision making by adding precision to patient care.

Finally, although these patient-facing applications are compelling, they represent a small fraction of this technology’s complete potential: as evidenced by the other five association tabs, there’s much more for Lasergene and Mastermind users to discover.