Thousands of mutations can exist in the DNA of cancer cells. Only a few of them, however, are responsible for cancer growth; the others are simply passengers.

Researchers may be able to identify better treatment targets if they can distinguish these damaging driver mutations from the neutral passengers.

An MIT-led team has developed a novel computer model that can quickly scan the whole genome of cancer cells and find mutations that occur more frequently than predicted, implying that they are driving tumor growth.

This form of prediction has proven difficult due to the high frequency of passenger mutations in some genomic areas, which obscures the signal of real drivers.

As explained by Maxwell Sherman, an MIT graduate student: “We created a probabilistic, deep-learning method that allowed us to get a really accurate model of the number of passenger mutations that should exist anywhere in the genome. Then we can look all across the genome for regions where you have an unexpected accumulation of mutations, which suggests that those are driver mutations .”

In this current study, researchers identified other changes throughout the genome that appear to contribute to tumor formation in 5 to 10% of cancer patients. According to the researchers, the findings could aid doctors in identifying meds that have a better likelihood of successfully treating those patients. At least 30% of cancer patients currently do not have a detectable driver mutation that can be used to guide treatment.

Since the sequencing of the human genome two decades ago, researchers have scoured the genome in search of mutations that contribute to cancer by allowing cells to expand uncontrolled or elude the immune system. This has resulted in the identification of targets such as the epidermal growth factor receptor (EGFR), which is often mutated in lung cancers, and BRAF, a melanoma driver. Specific treatments can now target both of these alterations.

Despite the fact that these targets have shown to be helpful, protein-coding genes comprise just roughly 2% of the genome. The remaining 98 percent contains mutations that can arise in cancer cells, but determining whether any of these mutations contribute to cancer growth has been significantly more challenging.

“There has really been a lack of computational tools that allow us to search for these driver mutations outside of protein-coding regions,” explains senior author Bonnie Berger. “That’s what we were trying to do here: design a computational method to let us look at not only the 2 percent of the genome that codes for proteins, but 100 percent of it.”

Researchers did this by searching cancer genomes for mutations that occur more often than expected by training a form of computational model known as a deep neural network. As a first step, they used genomic data from 37 different types of cancer to train the model. This gave the model enough information to figure out the background mutation rates for each type.

“The really nice thing about our model is that you train it once for a given cancer type, and it learns the mutation rate everywhere across the genome simultaneously for that particular type of cancer,” Sherman adds. “Then you can query the mutations that you see in a patient cohort against the number of mutations you should expect to see.”

The models were trained using data from the Roadmap Epigenomics Project and the Pan-Cancer Analysis of Whole Genomes, a worldwide data collection (PCAWG). The researchers were able to create a map of the expected passenger mutation rate across the genome using the model’s analysis, which allowed them to compare the expected rate in any group of areas (down to the single base pair) to the observed mutation count anywhere across the genome.

The landscape is shifting.

The model helped the MIT team expand the landscape of cancer-causing mutations. Currently, when tumors from cancer patients are examined for cancer-causing mutations, a known driver is found roughly two-thirds of the time. The latest findings from the MIT study suggest that an additional 5 to 10% of patients may have driver mutations.

“Cryptic splice mutations” are a type of noncoding mutation that researchers were interested in. Most genes are made up of long strings of exons and introns. Exons contain instructions for making proteins, and introns are spacers that are usually cut out of messenger RNA before it is turned into protein. There are cryptic splice mutations in introns, which can mess up the machinery in the cell that takes them out. As a result, introns are included when they aren’t supposed to be.

The researchers discovered that numerous cryptic splice mutations appear to affect tumor suppressor genes using their model. The tumor suppressors are spliced wrongly and stop acting when these mutations are present, and the cell loses one of its cancer-fighting defenses. The number of cryptic splice sites discovered by the researchers accounts for around 5% of the driver mutations discovered in tumor suppressor genes.

According to the researchers, targeting these mutations could provide a new strategy to potentially cure such patients. Short strands of RNA termed antisense oligonucleotides (ASOs) are used to patch over a mutated portion of DNA with the proper sequence in one proposed strategy that is currently under development.

“If you could make the mutation disappear in a way, then you solve the problem. Those tumor suppressor genes could keep operating and perhaps combat the cancer ,” Yaari explains. “The ASO technology is actively being developed, and this could be a very good application for it.”

Researchers also found a high number of noncoding driver mutations in the parts of some tumor suppressor genes that are not translated. It was already known that the faulty tumor suppressor gene TP53 accumulates several deletions in these sequences, known as 5′ untranslated regions. The MIT researchers discovered a similar pattern in the tumor suppressor ELF3.

Using this model, the researchers also investigated whether common mutations that were already known could possibly be responsible for different types of cancer. BRAF, which was previously associated to melanoma, was discovered to contribute to cancer progression in lesser percentages of other malignancies, such as pancreatic, liver, and gastric cancers, according to the researchers.

“That says that there’s actually a lot of overlap between the landscape of common drivers and the landscape of rare drivers. That provides an opportunity for therapeutic repurposing,” Sherman adds.

“These results could help guide the clinical trials that we should be setting up to expand these drugs from just being approved in one cancer, to being approved in many cancers and being able to help more patients.”

Image Credit: Getty

You were reading: Scientists identify the most aggressive cancer cells with new tool