The relapsed acute lymphoblastic leukemia network (ReALLNet): a multidisciplinary project from the spanish society of pediatric hematology and oncology (SEHOP).

Acute lymphoblastic leukemia (ALL) is the most common pediatric cancer, with survival rates exceeding 85%. However, 15% of patients will relapse; consequently, their survival rates decrease to below 50%. Therefore, several research and innovation studies are focusing on pediatric relapsed or refractory ALL (R/R ALL). Driven by this context and following the European strategic plan to implement precision medicine equitably, the Relapsed ALL Network (ReALLNet) was launched under the umbrella of SEHOP in 2021, aiming to connect bedside patient care with expert groups in R/R ALL in an interdisciplinary and multicentric network. To achieve this objective, a board consisting of experts in diagnosis, management, preclinical research, and clinical trials has been established. The requirements of treatment centers have been evaluated, and the available oncogenomic and functional study resources have been assessed and organized. A shipping platform has been developed to process samples requiring study derivation, and an integrated diagnostic committee has been established to report results. These biological data, as well as patient outcomes, are collected in a national registry. Additionally, samples from all patients are stored in a biobank. This comprehensive repository of data and samples is expected to foster an environment where preclinical researchers and data scientists can seek to meet the complex needs of this challenging population. This proof of concept aims to demonstrate that a network-based organization, such as that embodied by ReALLNet, provides the ideal niche for the equitable and efficient implementation of "what's next" in the management of children with R/R ALL.

DDQN-based optimal targeted therapy with reversible inhibitors to combat the Warburg effect.

We cover the Warburg effect with a three-component evolutionary model, where each component represents a different metabolic strategy. In this context, a scenario involving cells expressing three different phenotypes is presented. One tumour phenotype exhibits glycolytic metabolism through glucose uptake and lactate secretion. Lactate is used by a second malignant phenotype to proliferate. The third phenotype represents healthy cells, which performs oxidative phosphorylation. The purpose of this model is to gain a better understanding of the metabolic alterations associated with the Warburg effect. It is suitable to reproduce some of the clinical trials obtained in colorectal cancer and other even more aggressive tumours. It shows that lactate is an indicator of poor prognosis, since it favours the setting of polymorphic tumour equilibria that complicates its treatment. This model is also used to train a reinforcement learning algorithm, known as Double Deep Q-networks, in order to provide the first optimal targeted therapy based on experimental tumour growth inhibitors as genistein and AR-C155858. Our in silico solution includes the optimal therapy for all the tumour state space and also ensures the best possible quality of life for the patients, by considering the duration of treatment, the use of low-dose medications and the existence of possible contraindications. Optimal therapies obtained with Double Deep Q-networks are validated with the solutions of the Hamilton-Jacobi-Bellman equation.