Pushing cancer cells’ self-destruct button

“Synthetic lethality is an interaction between two genes where only inactivation of both genes kills a cell or organism.”

University of Otago PhD student Tom Kelly has been making heavy use of NeSI’s high performance computing platforms to investigate the possibility of using novel genetic techniques to help kill breast cancer cells. The technique he is investigating is called synthetic lethality.

“Synthetic lethality is an interaction between two genes where only inactivation of both genes kills a cell or organism.” Understanding this process and which gene pairs interact in this way can be modelled computationally.

Tom’s early work focused on a relatively small numbers of gene pairs. To scale this up to an entire genome, he required superior computing power. “I had developed a tool in R – and an R Shiny interface for execution by other biologists – to predict the synthetic lethal partners of a chosen query gene in a public cancer gene expression dataset. However, I became interested in the global genome-wide levels of synthetic lethality and genetic redundancy. Investigating wider numbers of candidate gene pairs and analysing the network is only feasible with access to high performance computing.”

He describes himself as an “inexperienced programmer”, but was surprised to find how accommodating the NeSI team has been. “Initially, I had very little experience or confidence using any command line interface apart from the R programming language. With support offered by NeSI and their patience with inexperienced programmers, scaling up my computational methods was straightforward and enabled me to answer biological research questions which could not be tackled without high performance computing.”

Researchers accessing NeSI services are supported in a nationally-consistent manner. “Online support from the NeSI team, including domain specific experts, makes me feel connected to a nationwide support network. That’s despite geographic isolation in Dunedin. NeSI hasn’t just assisted with technical problems either, but have also helped to tailor methodology to my research questions.”


Bioinformatic Analysis of Synthetic Lethality in Breast Cancer
Technical presentation delivered at eResearch NZ 2014

Before using NeSI, a naïve calculation for a job run was 10 days, 6 hours. Waiting two weeks is far too long to find out that there might be a bug in the code. With NeSI, Tom was able to speed that up 50 times. His runs have ended up taking just over 5 hours, using 64 cores.

To use 64 cores concurrently, Tom leveraged a technology known as MPI. MPI, or the Message Passing Interface, is the de facto standard for distributed programming in a large compute cluster. The first version of the standard was put in place in the early 90s and it continues to develop. While extremely effective at making use of many cores, MPI can be difficult to develop software for. Being able to support researchers through the learning and software engineering process is one of NeSI’s strengths.

“I highly recommend the experience of high performance computing in all fields, particularly the life sciences, not only for immediate benefits to researchers but as an opportunity for postgraduate students to learn how to use such resources with support offered by NeSI.” As NeSI develops its programme of workshops and other training initiatives, we hope to hear even more success stories like Tom’s. NeSI aims to support the capability development of all researchers, neutral of their research discipline or host institution.

Tom’s research is also heavily supported by access to open science data. The original gene pair datasets that Tom accessed were sourced from The Cancer Genome Atlas (TCGA Research Network, 2012) and the BC2116 Meta-Analysis dataset (Soon et al., 2011). These freely available datasets, collectively spanning tens of thousands of genes in each of the thousands of samples, enable this data-driven research. This requires the willingness of patients to donate their data for research into shared repositories of open datasets.

Open science continues with Tom’s own work. He has also made his code available for other biologists to facilitate their own research, via Shiny. Shiny is a tool that enables R code to be interacted with on screen via a web browser. It lowers barriers to entry and facilitates exploratory programming.

While the research is at an early stage, it is heartening to know that New Zealand has the research infrastructure to support work in this area. We wish Tom all the best as he continues his degree.

Next Case Study

Wahakura traditional Maori woven basket

Securely transferring health study data across the country

“The project team were originally going to courier hard drives across the country”