12:38pm Saturday 04 April 2020

Sifting out genetic defects from mounds of data

These are the optimistic future visions of Arnoldo Frigessi, Professor of Biostatistics at the University of Oslo.

For five years Professor Frigessi has led a group of researchers at the university’s Faculty of Medicine in their search for genes and “snips” that may be the root causes of disease in humans. Their work has received funding under the Research Council of Norway’s funding scheme for independent projects (FRIPRO).

Genetic mutations that cause disease

Professor Frigessi is now confident that his team has found what they have been looking for. They have identified not only one but perhaps as many as 25 snips – tiny mutations in genetic material. The term “snip” is coined from SNP, the acronym for single nucleotide polymorphism.

From among the one million snips in the human genome, Professor Frigessi and his team have been searching for the precise snips suspected of causing serious diseases in humans. The ultimate goal is to determine the risk of disease and then prevent its onset by repairing the genetic defects.

The professor and his colleagues are applying highly advanced statistics to the task.

Converting genetic data into biomedical knowledge

“As statisticians,” says Professor Frigessi, “we are not well-versed in molecular biology or medicine. But modern technology enables us to help medical researchers to analyse large volumes of data from which, on their own, they could not learn much.”

“Our work involves searching for genes or snips that may be causing disease. These genes, or parts of genes, are coded erroneously – mutated, in other words. To achieve results, we have carried out multiple experiments to find discrepancies between healthy people and sick people by comparing all their genes, roughly 30 000 per person. Genes that differ greatly between healthy and sick individuals are strong candidates for causing disease.”

But as the professor explains, not all their findings are accurate. “Converting genetic data into biomedical knowledge is a true challenge for us statisticians. Statistical methods are never 100 per cent certain. Not all data are perfect, and false interpretations or other factors may arise that can skew the results.”

False discovery rate

The Oslo statisticians have developed a method they call false discovery rate, which helps researchers to distinguish between genes with major deviations that are relevant and those which are false leads. This method has been applied in a breast cancer study that has formed the basis for the PhD dissertation of Hayat Mohammed, a student of the professor’s.

In her study, Ms Mohammed attempted to predict, using statistics, the effectiveness of radiation therapy to treat breast cancer. It is not certain that radiation therapy is beneficial or necessary in all cases. In fact, it is a treatment that often leads to many side effects, some major. Ms Mohammed’s study provides medical doctors with answers as to when this therapy can be particularly beneficial.

“Using a statistical method we call two-step lasso”, says Professor Frigessi, “we have pinpointed seven genes that interact successfully with the use of radiation therapy.”

Snips for facial recognition

Another topic of study is facial recognition, in which the research objective is to discover which snips are responsible for activating the parts of the brain we use to recognise faces. Professor Frigessi and his colleagues at Ullevål University Hospital took genetic data generated from patients and healthy individuals, and sorted this data into a million-voxel depiction of the brain to reveal which areas are associated with facial recognition and regulated by snips.

The data have been collected in an enormous database, and the result is quite clear: the Norwegian team has found 25 genes, or snips, that are important for recognising faces. The findings have been validated by new data from the US.

Medicines for genetic defects?

Professor Frigessi and his team are now forging ahead to develop and refine their methods for finding small effects hidden within large volumes of data.

“This is a long process, since every finding must be tested thoroughly to understand all the side effects and consequences, but I believe that in 10 years we will already be seeing the benefits of our research in the form of diagnostics for serious diseases caused by genetic defects. Ten years beyond that I envision personalised medicines on the market that can repair or compensate for individual genetic defects. This is what we hope our work in finding causal relationships using advanced statistics will lead to,” he says enthusiastically.

Professor Frigessi emphasises that statistics is a key tool of modern medical research. He believes that in addition to major investments in collecting data, it will be necessary to increase funding for statistical analysis, which is essential for gleaning knowledge from such large volumes of data.


Share on:

MORE FROM Genetics and Birth Defects

Health news