Improving the fight against cancer through AI
Cancer is one of the most complex diseases we face, both in terms of understanding and treatment. The leading experts of the 20th century simplified cancer research because they considered that this was the only way to develop useful scientific knowledge, being aware that in this way they would not be able to understand every factor that influences the appearance of cancer.
Fortunately, today we have a great ally in this fight: artificial intelligence. Thanks to artificial vision and machine learning (machine learning, comprising supervised learning; unsupervised learning; and reinforcement learning), data is processed in a completely new and efficient way that allows a better understanding of the disease and helps researchers to gain a deeper understanding than they could previously achieve.
To understand cancer and, in turn, treat it appropriately, there are many factors to consider:
Patient genetics: It is not only the complexity of the disease that complicates cancer treatment, but also the complexity of the patients. Everyone has their own genetic profile, and this genetic heterogeneity means that cancer treatments affect individuals from different genetic backgrounds in different ways. In other words, an effective treatment for someone of Eastern European ancestry might not be effective for someone of sub-Saharan African ancestry, and someone of East Asian ancestry might respond better to a completely different treatment plan.
Tumour heterogeneity: The different types of cells that make up an individual tumour, as well as the organ or tissue in which the cancer grows in the patient's body. All of this information can be difficult to pin down at any one time, but what makes it more challenging is that as cancer grows, its complexity tends to increase over time. As a result, it becomes more difficult to treat the cancer as it progresses. This is why it is so important to identify cancer early and destroy all tumour cells at once.
Patient lifestyle habits: Smoking, drinking, exercise or eating well all influence how cancer develops and how a patient may respond to treatment. There are many environmental and lifestyle factors that can play a crucial role in the development and treatment of cancer.
Patient medical history: The patient's entire medical profile, age, gender, previous illnesses, treatments and responses, as well as laboratory tests performed and results recorded add to the overall picture of the cancer and how to treat it.
All these factors can be summed up in one word: DATA. This is where AI brings out the full potential and helps researchers. By being able to work 24 hours a day and perform calculations much faster, we can obtain fundamental information that a few years ago required too much time from researchers. Thanks to machine learning algorithms, as we get more data we get better results, because these algorithms are getting "smarter" and more effective as they analyse it.
This allows patient data to be used in the selection of early-stage preclinical drugs, as we can cross-reference a patient's cancer data with their medical history, the history of cases with a similar cancer, and the drugs that have had the highest success rate. This enables better decision making and helps pharmaceutical companies and oncologists to predict treatment outcomes more accurately.
By contextualising data on drug formulations, background, patient lifestyle and various types of tumours and cancers, machine learning algorithms can easily identify the best drug formulations for specific types of cancer in defined patient populations; this information can be used in clinical trials, reducing the error rate and improving clinical trial outcomes in oncology.
This is what Predictive Oncology is working on, for example. Using a database of anonymised data from 150,000 patients, 131 tumour types and 30 types of cancers, they use machine learning algorithms to speed up the drug development process so that pharmaceutical companies can bring safe and effective drugs to market more quickly. This increases the likelihood of FDA approval.
It can also be used to administer chemotherapy and help healthcare providers develop optimal treatment plans. This can improve patient longevity and quality of life, as well as increase the likelihood that more patients will survive through remission and recovery.
In Spain, the Biomedical Genomics lab at IRB Barcelona has developed a computational method that identifies cancer-causing mutations for each type of tumour.
To find the mutations implicated in cancer, the scientists have relied on a key concept in evolution: positive selection. Mutations that favour the growth and development of cancer are found in higher numbers in different samples, compared to those mutations that occur randomly.
The proposed method learns from the data which attributes are distinctive for cancer-promoting mutations, providing useful information for the development of new therapeutic approaches.
This and other developments from the same lab aim to accelerate cancer research and provide tools for oncologists to choose the best treatment for each patient.
With AI as a tool, the idea of eliminating cancer is not such a distant hope. These tools offer us a new way to tackle one of the most complex diseases known.
