AI Tool Speeds Up Diagnosis & Treatment for Acute Myeloid Leukemia (AML)

Acute myeloid leukemia (AML), a rare and aggressive cancer affecting both blood and bone marrow, presents a significant challenge for clinicians. A new tool, developed by researchers at Northeastern University, offers a potential pathway to accelerate diagnosis and treatment planning for this complex disease. Currently, it can take a month or more after diagnosis for AML patients to begin receiving potentially life-saving treatment, a critical timeframe given the cancer’s poor prognosis – a median survival of less than five years after initial diagnosis.

The tool, recently awarded a patent, aims to drastically reduce this preliminary period. It’s designed to help oncologists not only diagnose AML but also map the diverse genetic mutations driving the disease in each individual patient. This detailed genetic profile is crucial, as AML doesn’t respond to a single treatment approach. Once these mutations are identified, the platform utilizes a computational model – a neural network – to suggest potential drugs and predict the likelihood of drug resistance.

Understanding the Genetic Complexity of AML

Kiran Vanaja, assistant research professor in bioengineering at Northeastern University, explains that understanding the internal workings of cancerous cells is a key hurdle in AML treatment. Researchers traditionally rely on techniques like gene sequencing and RNA sequencing to analyze these cells. However, Vanaja points out that a cell’s internal composition doesn’t always accurately reflect its behavior. Cancer cells, he explains, often lose their specialized function and revert to a more basic, stem-cell-like state.

To effectively test their new tool, Vanaja’s team approached the problem by considering genes as fundamental building blocks – akin to LEGOs. Each gene contributes to the overall expression of the cell, and the way these genes interact is incredibly complex. Even considering only the approximately 30,000 genes in the human genome, the possible combinations are vast. A deep learning neural network, inspired by the structure of the human brain, provides a means to sift through these combinations at the necessary speed.

This neural network operates through interconnected layers of processing nodes, similar to neurons. Each node analyzes a small component of the task and then directs the next step to another node in the subsequent layer. This allows the network to efficiently calculate the various ways a gene can be expressed by a cancer cell. More information on neural networks can be found at IBM’s explanation of the technology.

From Genotype to Phenotype: Bridging the Gap

Vanaja’s team built their AI model by feeding it data from thousands of cells from approximately a dozen AML patients. They then refined the model using data from numerous scientific studies on AML cells, enabling it to more accurately connect a cell’s genetic makeup (genotype) to its observable characteristics (phenotype). This represents particularly important because cancer cells often exhibit a mismatch between their genotype and phenotype as they desperately attempt to survive under therapeutic pressure.

In experiments, the team observed that AML cells undergoing treatment would “rewire” themselves, activating any possible survival mechanism. This rewiring creates a disconnect between the cell’s original genetic components and how it ultimately functions. The neural network aims to untangle these complex relationships and predict how a cell will respond to different treatments.

The potential impact of this tool is significant. Vanaja estimates that it could reduce the time from diagnosis to treatment recommendation from weeks to a single night. This accelerated timeline is critical, given the aggressive nature of AML. The tool isn’t limited to AML, however. Its core function – connecting genotype to phenotype – could be applied to other cancers and even other diseases.

The Role of Clinical Trials and Further Research

While the initial results are promising, further research is essential. The next step involves continued training of the model with additional patient samples and comparing its predictions against real-world clinical outcomes. Vanaja and his team also plan to explore the tool’s application to solid tumors, expanding its potential reach beyond blood cancers.

The development of this tool highlights the growing role of computational methods in cancer research. By leveraging the power of artificial intelligence, researchers are gaining new insights into the complex biology of cancer and developing innovative strategies to improve patient care. The patent awarded to Vanaja’s team, as reported by Northeastern University’s College of Engineering, can be found here.

For patients and families affected by AML, this research offers a glimmer of hope. While it’s important to remember that this tool is still under development, it represents a significant step forward in the fight against this deadly cancer. Individuals seeking more information about AML can find resources from organizations like the Leukemia & Lymphoma Society (https://www.lls.org/leukemia/acute-myeloid-leukemia).

Noah Lloyd is the assistant editor for research at Northeastern Global News and NGN Research.