Abstract:Around half of all cancer patients, world-wide, will receive some form of radiotherapy (RT) as part of their treatment. And yet, despite the rapid advance of high-throughput screening to identify successful chemotherapy drug candidates, there is no current analogue for RT protocol screening or discovery at any scale. Here we introduce and demonstrate the application of a high-throughput/high-fidelity coupled tumour-irradiation simulation approach, we call "GPU-GA", and apply it to human breast cancer analogue - EMT6/Ro spheroids. By analysing over 9.5 million candidate protocols, GPU-GA yields significant gains in tumour suppression versus prior state-of-the-art high-fidelity/-low-throughput computational search under two clinically relevant benchmarks. By extending the search space to hypofractionated areas (> 2 Gy/day) yet within total dose limits, further tumour suppression of up to 33.7% compared to state-of-the-art is obtained. GPU-GA could be applied to any cell line with sufficient empirical data, and to many clinically relevant RT considerations.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to screen and discover more effective treatment regimens through high - throughput and high - fidelity simulation methods in radiotherapy (RT). Although successful candidates for chemotherapeutic drugs can be identified through high - throughput screening, there is currently a lack of similar radiotherapy regimen screening or discovery methods. The paper introduces a high - throughput/high - fidelity coupled tumor - irradiation simulation method named "GPU - GA" and applies it to the study of EMT6/Ro spheres, a human breast cancer analogue. Specifically, by analyzing more than 9.5 million candidate regimens, the paper demonstrates the application effects of GPU - GA under two clinically relevant benchmarks, which are: 1. A fractionation regimen (BMI) of 1.25 Gy per fraction, twice a day. 2. A fractionation regimen (BMII) of 2 Gy per fraction, once a day. The research results show that the GPU - GA method can significantly improve the tumor - suppression effect under these benchmarks, increasing by 4.0% and 4.5% respectively compared with the existing high - fidelity/low - throughput computational search methods. In addition, by expanding the search space to higher - dose fractionation regimens (> 2 Gy/day), within the total dose limit, the tumor - suppression effect is further improved, up to 33.7%. In conclusion, the paper aims to efficiently explore the vast treatment - regimen search space through the GPU - GA method, thereby discovering more effective radiotherapy regimens and providing more promising candidates for clinical trials.

Identifying Promising Candidate Radiotherapy Protocols via GPU-GA in-silico