Revealing Human-growth Patterns in Transcriptomic Data
News May 19, 2014
Dr Adam Stevens, is a human development systems biologist working in Professor Peter Clayton's group at the UK's University of Manchester. In addition to industrially-funded and commercially-sensitive research into drug responses, Dr. Stevens and colleagues have published a number of papers on the pharmacogenomics of growth hormones.
Young humans grow in interesting ways. Dr. Stevens points out that humans are the only animals that return to rapid growth in puberty. Other animals - even other mammals that have a long juvenile phase - grow rapidly after birth and then their growth rate slows down until eventually tailing off as they reach their adult size.
Such observations go beyond quirky facts. Understanding how children grow is key to understanding many childhood diseases and conditions, particularly those affecting growth.
"An active return to growth is a biologically dangerous thing to do. If you want to understand the impact of drugs and disease on childhood you need to understand this," Dr. Stevens notes. "A child is very different at two than at four or 10 but medically they are treated the same."
His research aims to bring more understanding to the process. In addition to bench studies, he carries out in silico studies on a range of omics datasets derived primarily from blood samples throughout human childhood. "The problem is that there is no suitable animal model for this so to study the active return to long bone growth the only model we can use is humans."
Steven's datasets include transcriptomic data - which measures RNA and microarray RNA sequencing. He also looks at the integration of genetic data and omic data (such as proteomics and phosphor proteomic data).
He uses such datasets to study both normal childhood growth and growth with various impairments. For example, he looks at short stature, where there is a growth hormone deficiency, and chromosomal syndromes, for example Turner syndrome, where girls are often treated with growth hormones. He also looks at diseases in children where growth has an impact, such as cancers. "Many cancers in children have a specific onset age, which suggests that different sets of genes are involved," he explains. Some drugs, such as steroids, also affect growth.
"Now we understand better how to map growth from a molecular level and its impact on pharmaco genetics," explains Dr. Stevens. "The methods we've developed allow us to study it more clearly, with highly-resolved molecular maps."
To assist with this research, he turned to the Qlucore Omics Explorer software from Sweden-based Qlucore. Dr. Stevens found that a fundamental problem when running a project was that you might need a statisticians but they don't always understand the biology. One of the key drivers in using the software was the visually-driven interface. Data needed to be presented in a way that could be understood and analysed by reviewers who are not experts in statistics and mathematics.
"The sheer genius is letting the visualisation drive the analysis. We want to have software that's accessible to non-statisticians. A fundamental problem that I see all the time is if you're running a project and realise you need statistics expertise you find statisticians but they don't understand the biology. There is a communication problem."
Dr. Stevens imports different microarrays in a range of conditions into the tool and tests whether there are differences or not. The industry-standard transcriptomic datasets might include around 400 microarrays, each with 55,000 data points. He has also imported other omics data into Qlucore. These datasets, he says, tend to be smaller but still have thousands of data points per sample.
What he does with this data depends on the data and the question he wants to ask.
"If I come in with a hypothesis, it's simple to use and replicates what lots of other tools do," he notes. "The power of Qlucore is in assessing structure when I'm coming in with hypothesis-free data."
In fact, he observes that several times he has been able to spot previously undiscovered patterns also in published datasets because of the visualisation capabilities of the software.
There are hundreds of thousands of structural variants (SVs) in the human genome that are hard to identify than single nucleotide variants. Researchers have used a full suite of genomic technologies to analyze the number of SVs in three individuals, and the results present as the most comprehensive catalog of SVS to date.READ MORE