UBC statistician John Petkau employs statistics and MRI imaging to advance multiple sclerosis treatments.
It’s not just a numbers game—it’s translation.
We’ve tackled many of the storage and processing problems inherent in an increasingly data-rich society. But making sense of all the genomic data, social network interactions and customer records is the real trick.
Statisticians at UBC are tasked with sifting through that data to help find the answers fellow researchers and industry partners are looking for. Not only can the data be highly complex, but understanding how another discipline works and what matters to scientists in that discipline is part of the process.
“Communication is very important,” says statistician John Petkau. “You need to build a personal relationship with people in other units, and that takes time. In science we tend to be very discipline specific, but statistical research often involves a great deal of collaboration.”
That means statisticians can be pulled into linguistics, computer science, wood product engineering, even marine mammal research. Petkau’s focus is medical research.
Petkau was one of the first statisticians to study connections between air pollution and human health—applying statistics to the issue in the late 80s. Petkau has also worked to improve the evaluation of multiple sclerosis (MS) treatments, creating models that allow software to flag patients who are experiencing an unusual increase in MS lesions, a signal of increased risk of imminent clinical worsening.
Petkau is also working on more 'speculative' work. There are currently no MS treatments that repair the damage the disease causes to nerve cells. But researchers are interested in developing new regenerative treatments, and if they do, they need a way to evaluate the results.
Working with researchers at the United States National Institutes of Health, Petkau has developed a way to evaluate potential treatments by analyzing measurements of the individual voxels — voxel combines volume and pixel, and is applied to three-dimensional imagery — from images obtained through MRIs.
“You can measure the intensity of signals from each one of those little voxels that corresponds to a lesion each month, and potentially use that information to assess whether the lesion is repairing itself over time.”
Petkau’s method could allow for smaller, yet still reliable, clinical trials—what he calls the “holy grail” of health research.
Gabriela Cohen Freue. Source: UBC Science.
Cohen Freue's students present results of real case studies which involved collaborations with other units. Source: UBC Science.
The usual suspects
Like Petkau, Gabriela Cohen Freue's focus is the medical sciences. She is developing methods that can help predict organ rejection using genetic information.
“When I talk with people they don’t understand what I do,” Cohen Freue says with a smile. “They think statistics has to do with determining income or producing statistical summaries. They’re surprised when I tell them I work with doctors to understand the data they have and to use the data to improve the health treatment of patients.”
Cohen Freue works on statistical methods which can help medical professionals understand the complexity of genomics data and its variables.
“If your doctor thinks your kidney isn’t working well they’ll do an analysis of your creatinine to see if it’s above or below normal ranges,” she says. “But for other diseases, like cancer, there is nothing like this.”
Doctors often can’t diagnose cancer until signs are evident, and that means it is hard or impossible to reverse the disease’s progression. Cohen Freue attempts to “catch” the disease early in the act, diagnosing it sooner by finding biomarkers.
“The blessing and the curse of genomics data is the large number of variables that are measured. Too many genes are expressed in human blood, which, for a statistician means too many variables to be analyzed. Some of these variables are louder than others, some ‘speak’ in groups, some contain outlying measures,” Cohen Freue’s says.
Cohen Freue is working to develop robust statistical methods to select the most relevant variables to predict patients’ outcomes. She looks at the rich information contained in patients’ genes to develop minimally invasive blood tests that can help diagnose and predict a disease or organ rejection.
“The human body is complex. It is difficult to measure, read and translate its message,” she says.
Matías Salibián-Barrera. Source: UBC Science.
UBC statistics graduate student Yang Liu says hello to a sea lion. Source: UBC Marine Mammal Research Unit.
Diving into data
We tend to think more is better. And certainly more data can be helpful. But as data collection and storage has become easier, researchers can find themselves bobbing up and down in a sea of information.
Animal tag sensors can transmit the position of marine mammals and record the water properties of the areas where they are swimming. But with sensors recording data several times a second, the amount of data adds up, becoming monumental. Statistician Matías Salibián-Barrera and his colleagues have been working with UBC’s Marine Mammal Research Unit to find ways to recreate more accurate marine mammal swimming paths and to process the information faster. The water property data could even be mined for climate change studies.
But just because technology has improved the ease of data collection doesn’t mean data is collected systematically. A smartphone can provide information on barometric pressure and could be used to crowdsource micro-scale weather data. Sounds great, right? There’s a caveat.
“Unlike static sensors like weather stations, people walk around as they take these measurements are taken and their movements may not be random. We visit some places more than others,” Salibián-Barrera says.
It’s just the sort of issue statisticians need to grapple with, because vast repositories of data don’t equal accurate information, let alone new knowledge.
“The data can’t speak for itself,” Salibián-Barrera says. “We have to evaluate it and translate it.”
Support the Statistics Fund for Excellence, which allows the Department of Statistics to assist graduate and undergraduate students.
Need some stats help? The Statistical Consulting and Research Laboratory (SCARL) operated by the Department of Statistics offers consulting services to all faculties as well as to off-campus clients.
Thanks to the MS/MRI Research Group at the Djavad Mowafaghian Centre for Brain Health (DMCBH) for their assistance in the writing of this piece.