Abstract Geometric Semantic Genetic Programming ( $$\mathsf {GSGP}$$ ) is a powerful variant of Genetic Programming (GP) that defines genetic operators inducing unimodal fitness landscapes. In recent years, a new mutation operator, Geometric Semantic Mutation with Local Search (GSM-LS), has been proposed to include a local search step in the mutation process. The core idea of GSM-LS is to incorporate a linear regression step during mutation, thereby accelerating convergence toward high-quality solutions. While GSM-LS helps the convergence of the evolutionary search, it is prone to overfitting. Thus, it was suggested to apply GSM-LS only for a limited number of generations and then revert to standard geometric semantic mutation. A more recently defined variant of $$\mathsf {GSGP}$$ (called $$\mathsf {GSGP}$$ -reg) also includes a local search step, but shares similar strengths and weaknesses with GSM-LS. Here, we investigate several strategies to mitigate overfitting in GSM-LS and $$\mathsf {GSGP}$$ -reg, ranging from simple regularized regression techniques to adaptive methods that estimate overfitting risk at each mutation. The latter approaches partition the training set into two subsets: one used to perform the mutation, and the other to evaluate the risk of overfitting based on the mutation’s impact on held-out data. Experimental evaluations across seven real-world regression benchmarks show that, while plain GSGP underperforms on all datasets, methods incorporating local search often achieve significantly better test performance. For example, on the Airfoil dataset, the GSM-LS variant achieves a median RMSE below 10 compared to 30 with standard GSGP. On the LD50 and Bioavailability datasets, the proposed gen and ridge-regularized variants effectively mitigate overfitting, reducing test RMSE by up to 40% relative to baseline GSGP. We conclude that local search, when used with regularization strategies, enhances GSGP’s performance and generalization capability across a diverse range of tasks.

Despite their impressive performance in classification tasks, neural networks are known to be vulnerable to adversarial attacks, subtle perturbations of the input data designed to deceive the model. In this work, we investigate the correlation between these perturbations and the implicit bias of neural networks trained with gradient-based algorithms. To this end, we analyse a representation of the network’s implicit bias through the lens of the Fourier transform. Specifically, we identify unique fingerprints of implicit bias and adversarial attacks by calculating the minimal, essential frequencies needed for accurate classification of each image, as well as the frequencies that drive misclassification in its adversarially perturbed counterpart. This approach enables us to uncover and analyse the correlation between these essential frequencies, providing a precise map of how the network’s biases align or contrast with the frequency components exploited by adversarial attacks. To this end, among other methods, we use a newly introduced technique capable of detecting nonlinear correlations between high-dimensional datasets. Our results provide empirical evidence that the network bias in Fourier space and the target frequencies of adversarial attacks are highly correlated and suggest new potential strategies for adversarial defence. Code is available at https://github.com/lorenzobasile/ImplicitBiasAdversarial

Single-cell RNA and ATAC sequencing technologies enable the examination of gene expression and chromatin accessibility in individual cells, providing insights into cellular phenotypes. In cancer research, it is important to consistently analyze these states within an evolutionary context on genetic clones. Here we present CONGAS+, a Bayesian model to map single-cell RNA and ATAC profiles onto the latent space of copy number clones. CONGAS+ clusters cells into tumour subclones with similar ploidy, rendering straightforward to compare their expression and chromatin profiles. The framework, implemented on GPU and tested on real and simulated data, scales to analyse seamlessly thousands of cells, demonstrating better performance than single-molecule models, and supporting new multi-omics assays. In prostate cancer, lymphoma and basal cell carcinoma, CONGAS+ successfully identifies complex subclonal architectures while providing a coherent mapping between ATAC and RNA, facilitating the study of genotype-phenotype maps and their connection to genomic instability.

BACKGROUND: The combined effects of biological variability and measurement-related errors on cancer sequencing data remain largely unexplored. However, the spatio-temporal simulation of multi-cellular systems provides a powerful instrument to address this issue. In particular, efficient algorithmic frameworks are needed to overcome the harsh trade-off between scalability and expressivity, so to allow one to simulate both realistic cancer evolution scenarios and the related sequencing experiments, which can then be used to benchmark downstream bioinformatics methods. RESULT: We introduce a Julia package for SPAtial Cancer Evolution (J-SPACE), which allows one to model and simulate a broad set of experimental scenarios, phenomenological rules and sequencing settings.Specifically, J-SPACE simulates the spatial dynamics of cells as a continuous-time multi-type birth-death stochastic process on a arbitrary graph, employing different rules of interaction and an optimised Gillespie algorithm. The evolutionary dynamics of genomic alterations (single-nucleotide variants and indels) is simulated either under the Infinite Sites Assumption or several different substitution models, including one based on mutational signatures. After mimicking the spatial sampling of tumour cells, J-SPACE returns the related phylogenetic model, and allows one to generate synthetic reads from several Next-Generation Sequencing (NGS) platforms, via the ART read simulator. The results are finally returned in standard FASTA, FASTQ, SAM, ALN and Newick file formats. CONCLUSION: J-SPACE is designed to efficiently simulate the heterogeneous behaviour of a large number of cancer cells and produces a rich set of outputs. Our framework is useful to investigate the emergent spatial dynamics of cancer subpopulations, as well as to assess the impact of incomplete sampling and of experiment-specific errors. Importantly, the output of J-SPACE is designed to allow the performance assessment of downstream bioinformatics pipelines processing NGS data. J-SPACE is freely available at: https://github.com/BIMIB-DISCo/J-Space.jl .

In this work we investigate, by means of a computational stochastic model, how tumor cells with wild-type p53 gene respond to the drug Nutlin, an agent that interferes with the Mdm2-mediated p53 regulation. In particular, we show how the stochastic gene-switching controlled by p53 can explain experimental dose-response curves, i.e., the observed inter-cell variability of the cell viability under Nutlin action. The proposed model describes in some detail the regulation network of p53, including the negative feedback loop mediated by Mdm2 and the positive loop mediated by PTEN, as well as the reversible inhibition of Mdm2 caused by Nutlin binding. The fate of the individual cell is assumed to be decided by the rising of nuclear-phosphorylated p53 over a certain threshold. We also performed in silico experiments to evaluate the dose-response curve after a single drug dose delivered in mice, or after its fractionated administration. Our results suggest that dose-splitting may be ineffective at low doses and effective at high doses. This complex behavior can be due to the interplay among the existence of a threshold on the p53 level for its cell activity, the nonlinearity of the relationship between the bolus dose and the peak of active p53, and the relatively fast elimination of the drug.

This paper outlines the major components and function of the technologically integrated oncosimulator developed primarily within the Advancing Clinico Genomic Trials on Cancer (ACGT) project. The Oncosimulator is defined as an information technology system simulating in vivo tumor response to therapeutic modalities within the clinical trial context. Chemotherapy in the neoadjuvant setting, according to two real clinical trials concerning nephroblastoma and breast cancer, has been considered. The spatiotemporal simulation module embedded in the Oncosimulator is based on the multiscale, predominantly top-down, discrete entity-discrete event cancer simulation technique developed by the In Silico Oncology Group, National Technical University of Athens. The technology modules include multiscale data handling, image processing, invocation of code execution via a spreadsheet-inspired environment portal, execution of the code on the grid, and the visualization of the predictions. A refining scenario for the eventual coupling of the oncosimulator with immunological models is also presented. Parameter values have been adapted to multiscale clinical trial data in a consistent way, thus supporting the predictive potential of the oncosimulator. Indicative results demonstrating various aspects of the clinical adaptation and validation process are presented. Completion of these processes is expected to pave the way for the clinical translation of the system.

BACKGROUND: Anti-tumor therapies aim at reducing to zero the number of tumor cells in a host within their end or, at least, aim at leaving the patient with a sufficiently small number of tumor cells so that the residual tumor can be eradicated by the immune system. Besides severe side-effects, a key problem of such therapies is finding a suitable scheduling of their administration to the patients. In this paper we study the effect of varying therapy-related parameters on the final outcome of the interplay between a tumor and the immune system. RESULTS: This work generalizes our previous study on hybrid models of such an interplay where interleukins are modeled as a continuous variable, and the tumor and the immune system as a discrete-state continuous-time stochastic process. The hybrid model we use is obtained by modifying the corresponding deterministic model, originally proposed by Kirschner and Panetta. We consider Adoptive Cellular Immunotherapies and Interleukin-based therapies, as well as their combination. By asymptotic and transitory analyses of the corresponding deterministic model we find conditions guaranteeing tumor eradication, and we tune the parameters of the hybrid model accordingly. We then perform stochastic simulations of the hybrid model under various therapeutic settings: constant, piece-wise constant or impulsive infusion and daily or weekly delivery schedules. CONCLUSIONS: Results suggest that, in some cases, the delivery schedule may deeply impact on the therapy-induced tumor eradication time. Indeed, our model suggests that Interleukin-based therapies may not be effective for every patient, and that the piece-wise constant is the most effective delivery to stimulate the immune-response. For Adoptive Cellular Immunotherapies a metronomic delivery seems more effective, as it happens for other anti-angiogenesis therapies and chemotherapies, and the impulsive delivery seems more effective than the piece-wise constant. The expected synergistic effects have been observed when the therapies are combined.