An artificial intelligence-based model has been used to match molecularly targeted agents (MTAs) to individual molecular tumour profiles, in the hope of improving the clinical benefit of precision oncology.
Precision oncology is the molecular profiling of tumours to identify individual targetable alterations. The practise is rapidly developing and has begun to enter mainstream clinical practice. To date, efforts in precision oncology have been largely focussed on predicting the response of a molecularly targeted agent (MTA) following the absence or presence of a genomic alteration.
Mutations of driver genes are what cause tumour growth, and biomarkers are structures or processes that can be measured to predict disease. Precision oncology pairs MTAs with a pre-defined driver gene or a single biomarker. However, each tumour is thought to harbour an average of 3-4 genetic driver alterations, discounting the non-coding regions, and driver genes can be associated with multiple MTAs within the same tumour. This means that the use of single biomarkers presents severe limitations for precision oncology and is restricting the success rate of the approach.
Due to advances in molecular diagnostics, particularly next-generation sequencing (NGS), it is now possible to analyse multiple driver genes in parallel, instead of one at a time. However, precision oncology still faces complex challenges. For example, assessing the significance of all the detected genetic alterations of all the potential driver genes. This is then followed by the task of choosing the correct target and matching it with an effective MTA. On top of that, the different sensitivity of individual genetic alterations to different MTAs needs to be considered.
Using artificial intelligence to predict response of MTAs
Researchers have recently developed a computational method, assisted by artificial intelligence (AI), that prioritises potential MTAs based on the individual molecular profile of a tumour. The system is called the digital drug-assignment (DDA). It can search through a database that consists of over 12,000 driver-target-MTA interactions, to automatically highlight the MTAs that are related to a patients complete molecular profile.
To evaluate the model’s clinical benefit, the scientists uploaded data from 113 precision oncology patients into the DDA. All the patients were part of a clinical trial called SHIVA01 and had been treated with MTAs matched to individual genetic alterations of their tumour. The DDA then assigned a score to each MTA. It was found that the MTA scores increased by three-fold for patients in the disease control group compared to patients with progressive disease. Essentially, this suggests that MTAs with higher DDA scores were associated with a considerably higher clinical benefit.
Prospects of the DDA
These results indicate that the AI-based DDA was capable of prioritising MTAs for successful precision oncology. The data suggests that choosing an MTA based on a higher DDA score led to improved clinical benefit. This is hugely exciting as the AI-based approach may be the first step in potentially overcoming the limitations of single biomarker-based decisions in precision oncology, ultimately paving the way for personalised tumour treatment strategies.
Further studies to evaluate the absolute clinical benefit of using MTAs based on DDA scores are required before the introduction of new treatment options, purely according to the scoring system, are rolled out. Nevertheless, MTAs and corresponding DDAs could be co-developed in the future, leading to the acceleration, reduced cost and decreased risk of drug development.
Ultimately, it is hoped that the predictive scores generated by these computational methods will eventually help to make better treatment decisions in the upcoming digital age of precision oncology.
Image credit: FreePik romanzaiets