Computer Science and Mathematics

Abstract: Spatiotemporal models of chimeric antigen receptor T-cell therapy can clarify how interacting biological mechanisms shape therapeutic response, but global sensitivity analysis and response-regime mapping require many computationally expensive simulations. We developed an adaptive surrogate-assisted algorithm that combines these tasks within a single computational framework. Independent Gaussian process surrogates were constructed for the final and minimum normalised tumour burdens, and new simulations were selected according to surrogate uncertainty and proximity to regime boundaries. Surrogates trained on 120 simulations achieved normalised root-mean-square errors of 0.00148 and 0.01058 on a fully held-out test set. Regime classification agreed with direct simulations at all 30 test points and 12 additional points near regime boundaries. The stopping criterion was first satisfied after 84 training simulations, while the advantage of adaptive sampling was most evident at intermediate computational budgets. Antigen escape and cytotoxicity dominated the variance in final tumour burden, with total-order Sobol indices of 0.4944 and 0.4382, respectively. Chemotaxis contributed much less to global variance but shifted the boundary between partial response and regrowth. These results show that global sensitivity ranking and regime-boundary analysis provide complementary information and can be combined within a limited budget of direct full-model simulations.

Abstract: Translational biomedical research is very important to the health of cancer patients. The development of novel biomedical therapies requires continuous iteration between computational prediction and experimental validation. However, the infrastructure for translational biomedical research is still under development. Currently, researchers face significant bottlenecks in two critical areas: first, the visualization of dynamic biological processes from limited experimental data, and second, the integration of diverse analytical tools required for data-driven decision-making. This project presents two complementary engineering solutions: a deep learning model for spatiotemporal biological prediction, and an integrated web-based analytics platform. The deep learning model employs a novel Convolutional LSTM (ConvLSTM) architecture trained on 10,224 microscopy images to predict bacterial infection dynamics in colorectal cancer organoids, achieving a Peak Signal-to-Noise Ratio (PSNR) of 17.3 dB and Structural Similarity Index (SSIM) of 0.59. These metrics indicate successful capture of macro-level morphological changes while identifying clear pathways for improvement. The analytics platform consolidates four machine learning modules (dose-response prediction, time-series forecasting, database querying, and polynomial regression) into a single interface that delivers results in under one second without requiring programming knowledge. In validation testing, the platform reduced experimental screening time from weeks to minutes, representing a 60-fold workflow acceleration. By reducing complex computational analysis to single-click operations, this work lowers the barrier between experimental biologists and machine learning tools.

Abstract: Large language models (LLMs) achieve strong reasoning performance through Chain-of-Thought (CoT), yet their autoregressive generation provides limited supervision over the correctness of intermediate steps. We introduce Step-wise Check Optimization for Process-Enhanced Reinforcement Learning (SCOPE-RL), a reinforcement learning framework that equips reasoning steps with executable mathematical checks. Each step is associated with a typed verification operator, such as symbolic equivalence, substitution, boundary evaluation, or range checking. Candidate checks are certified using reference solutions and controlled counterexamples to ensure both validity and discriminative ability. During reinforcement learning, these checks are executed directly on intermediate reasoning states to provide a process-level reward, which is combined with the final-answer reward under the standard GRPO framework. Experiments across multiple language models and mathematical reasoning benchmarks show that SCOPE-RL consistently improves accuracy over strong RLVR baselines while producing more verifiable and mathematically consistent reasoning trajectories.

Abstract: Following Livadiotis G. and McComas D. J. (2023) [1], we propose a new type of single-stand (ss)DNA/RNA frameshift mutations that occur spontaneously due to information exchange between the ssDNA sequence of length bases (N) and the mutation sequence of length bases (M), and respect the kappa-addition symbol ⊕κ. We call these proposed mutations Kappa-Frameshift Background (KFB) mutations. We find that entropy defects originate in the interdependence of the information length systems (or their interconnectedness, that is, the state in which systems with a significant number of constituents (information length bases) depend on, or are connected with each other by the proposed KFB-mutation). We also quantify the correlation among ssDNA/RNA information length bases (N) and (M) due to information exchange. In the presence of entropy defects, the Landauer’s bound and minimal metabolic rate for a biological system is modified. We observe that the different N and κ scales are manifested in the double evolutionary emergence of the proposed biological system through subsystems correlations. For specific values of the kappa parameter we can expect deterministic laws associated with a single biological polymer in the short term; over time, before the polymer explores all the possible ways it can exist.

Abstract: Virtual-patient models need a way to represent a person’s biological state before they can model how that state may change. Many AI-for-science models focus on molecules, cells, or a single organ, while treatment decisions are made for people who may have several conditions and many possible interventions. We used drug repurposing as a human-scale test of whether organized biological modules can help rank known drug–disease relationships.We introduce SteeraMed Bench, which compares panels drawn from a 332-module library across drug-repurposing tasks. The benchmark included 1,916 small molecules and five primary chronic-disease tasks, with an exploratory analysis across 23 disease categories. In standard cross-validation, recall@20—the fraction of known positive drug–disease pairs found among the top 20 ranked drugs—was 0.494 for the 117-module nutraceutical panel across five diseases, close to 0.524 for the full library. In the 23-category analysis, the full library ranked first in only 9 categories; the best-performing panel varied by disease: aging-hallmark panels led for type 2 diabetes and osteoporosis, food-as-medicine for depression, and nutraceutical modules for atherosclerosis/hyperlipidemia. These findings show that different biological panels carry different ranking signals within the benchmark. Performance fell when drugs with related target families were kept out of training, and was near chance when an entire disease was held out.We also tested whether a language-model-assisted agent could propose new modules for the library. Two of 14 proposals showed nominal positive gains in the second round, but neither remained significant after correction for testing seven concepts. SteeraMed Bench provides a shared way to compare and update biological representations for future virtual-patient research. It is a benchmark for representation and candidate prioritization, not a complete virtual patient or a validated drug-discovery system.

Abstract: In this work, Bayesian estimation is performed for the parameters of a mathematical model, formulated as a system of ordinary differential equations, that describes cannibalism dynamics in a larval populations of Atractosteus tropicus were reared in a pond. The model divides the population into three subclasses: vulnerable, cannibalistic, and non-vulnerable individuals. Parameter estimation is conducted under two scenarios: a fed population and an unfed population. Correlation matrices are calculated for both cases. For the fed population, the most influential parameters are food quality and the time devoted to managing cannibalistic individuals. In contrast, for the unfed population, the strongest correlation occurs between the mortality rate and the transition rate of cannibalistic individuals to the non-vulnerable class. Therefore, providing larvae with high-quality food substantially reduces cannibalism and the handling time required and, consequently, increases the survival of the species.

Abstract: This paper combines hyperstructure theory with fuzzy soft sets to develop a parameterized algebraic model for epigenetic inheritance. By defining hyperoperations dependent on environmental inputs, we model external factors such as temperature fluctuations, soil nutrients, and acoustic stress. We study the structural properties of these systems and show that the resulting fuzzy soft H_v-semigroups preserve weak associativity.

Abstract: As microbiome research increasingly seeks to identify true ecological shifts, transitioning from associational to causal approaches is essential. However, detecting structural changes across independent networks remains challenging due to the absence of established biological ground truths and the small, imbalanced sample sizes typical of microbiome cohorts. To address this, we extend an existing network comparison framework to enable node-level mechanism-shift detection under the direct linear non-Gaussian acyclic model (DirectLiNGAM). We evaluate the Naive, Bootstrap, and Relative sample size Bootstrap Stability (RSBS) estimators across extensive synthetic discovery runs and a semi-synthetic, batch-corrected human gut microbiome cohort. Our results demonstrate that resampling-based estimation consistently outperforms a single-fit Naive baseline by trading marginal recall for substantial precision gains. On both semi-synthetic and synthetic data, standard Bootstrap is optimal for comparing datasets of equal size, whereas RSBS is the only estimator that reliably handles imbalanced cohorts. This advantage strengthens as network dimensionality increases and persists under authentic compositional noise. Navigating this complex and emerging research area is currently constrained by limitations in data quantity and quality, scarce biological knowledge, and a lack of dedicated software. To address these critical gaps, we provide the complete benchmark pipeline and synthetic data generators as an open-access Python package, causal-comparator, to support node-level mechanism-shift detection across systems biology applications.

Abstract: Motivation: Multiomics data integration is essential for understanding complex biological systems, yet existing network-based approaches operate at a single level—either patient similarity networks or molecular interaction networks—without bridging the two. We address the gap between patient-level integration and molecular-level network analysis by proposing a hierarchical framework that unifies both. Results: We introduce HiNoN (Hierarchical Network-of-Networks), a two-level framework in which a fused patient-similarity network is linked to subgroup-specific microbial association and host-gene co-expression networks. Across 200 simulation runs, subgroup recovery was strongest for large cohorts (mean ARI 0.95 at n=500) and two-group settings (mean ARI 0.88), but declined with increasing noise and subgroup number. We then performed leakage-free repeated nested cross-validation on 50 paired coronary artery disease samples. Under the prespecified 16-component representation, early concatenation outperformed HiNoN-path2-SGC in ROC–AUC (0.923 versus 0.911), PR–AUC (0.958 versus 0.949), and calibration. Sensitivity analysis showed a different pattern under aggressive compression: with eight components, HiNoN-path2-SGC improved ROC--AUC by 0.073 and also improved balanced accuracy, Brier score, and calibration error. Unsupervised clusters had negligible agreement with disease labels (ARI =-0.006; NMI =0.016) and are therefore treated as exploratory. These results characterize HiNoN as a topology-aware regularizer under representation scarcity, rather than a uniformly superior classifier.

Abstract: Since 1995, the intake of individual nutrients has been evaluated using fuzzy logic, with aggregation of the resulting fuzzy sets enabling the weighing of one nutrient against another, and the evaluation and optimization of overall nutrient intake. The aggregation operator used in earlier work, defined as the product of the minimum and harmonic mean operators (MIN×H), performed well in practice but had undesirable mathematical properties: it is not associative, model extensions can shift results even when all conditions are fully satisfied, and the resulting evaluation field contains discontinuities inconsistent with biological behavior. We therefore reviewed the family of parametrized T-norms for an alternative and identified the Schweizer (3) operator, fitting its parameter to best match the existing operator using 254 real nutrition protocols from a sustainability study; strong agreement was achieved at a parameter setting of p = 6. The new operator satisfies all required mathematical properties and produces a smooth, simply structured evaluation field that reveals synergistic and substitution effects among foodstuffs, as well as the essentiality of certain foodstuffs under specific conditions. The resulting evaluation field shows a wide, high-dimensional plateau rather than a sharp peak, interpretable as stability, error tolerance, and a high degree of diversity, offering a new, mathematically robust perspective on healthy nutrition and the interactions among foodstuffs.

Abstract: Machine learning has become a central framework for analyzing high-dimensional biological data generated by genomics, transcriptomics, proteomics, single-cell sequencing, and multi-omics technologies. However, predictive modeling is frequently limited by severe class imbalance, where biologically important populations including rare disease subtypes, uncommon molecular states, and low-frequency cell populations are underrepresented. Synthetic oversampling addresses this challenge by increasing minority-class representation, with the Synthetic Minority Oversampling Technique (SMOTE) being the most widely used approach. However, conventional SMOTE assumes that geometric proximity reflects biological similarity, an assumption often violated in high-dimensional, nonlinear, and heterogeneous biological data. This review examines clustering-guided synthetic generation methods that integrate unsupervised learning before oversampling to preserve underlying biological structure. We evaluate partition-based, density-aware, fuzzy, cluster-filtering, and representation-based approaches, emphasizing their mathematical foundations, computational assumptions, and applications in molecular classification, rare population analysis, single-cell modeling, and multi-omics integration. Key challenges including cluster instability, preservation of rare biological populations, synthetic data validation, and distinguishing statistical rarity from biological significance are discussed. Finally, we outline emerging directions involving biological foundation models, graph-based learning, multi-omics representation spaces, and biologically constrained generative frameworks. These advances represent a shift from conventional data balancing toward structure-aware synthetic modeling that preserves meaningful biological organization while improving predictive performance.

Abstract: Cancer-associated alterations perturb molecular systems whose organization is only partially captured by protein–protein interaction topology. We asked whether Gene Ontology (GO)- derived semantic information can improve reconstruction of mutation-defined cancer modules and, critically, whether the magnitude and ontology source of this improvement vary across tumour types. We constructed recurrent non-silent somatic-mutation modules for 31 non-haematological The Cancer Genome Atlas (TCGA) projects and embedded them in a frozen human interaction network enriched with a nine-dimensional GO semantic representation. Topology-only random walk with restart (RWR) was compared with semantic-RWR under symmetric leave-one-out module reconstruction. The benchmark comprised 7,440 evaluations spanning 31 cancers, module sizes of 50, 200 and 400 proteins, intact or 20% edge-depleted interactomes, two graph perturbation seeds, five module-subsampling seeds, and four semantic configurations (combined, Biological Process, Molecular Function and Cellular Component). Combined semantic diffusion significantly improved NDCG@100 in 27 of 31 cancers after Holm correction. Semantic gain increased strongly with module size and was largely preserved after removal of 20% of interactome edges. The optimal ontology branch differed among cancers: Molecular Function was optimal in 13, Biological Process in 12, Cellular Component in 4, and the combined representation in 2. These findings indicate that cancer modules are not organized by topology alone and support the existence of tumour-specific semantic network architectures: systems-level patterns describing whether functional coherence is expressed primarily through biological processes, biochemical activities, cellular compartments, or combinations thereof. Biologically, this heterogeneity is consistent with the known convergence of diverse cancer mutations onto shared pathways, complexes and cellular programs. PanCancerSemNet therefore provides a framework for interpreting cancer-associated alterations as context-dependent perturbations of functionally annotated molecular systems rather than as isolated mutated genes.

Abstract: We propose a mathematical model for the structure of consciousness, and we do not treat the feeling itself. We build the model on a phase-augmented Recursive Heaviside Sequence Function, a cascade of nested smooth step functions in which each layer is one episodic memory carrying a time threshold \(\tau_i\), a sharpness \(s_i\), and an emotional phase \(\theta_i\). We propose that the objective of consciousness is survival, the drive to keep the remaining distance to a fixed endpoint short. Our central claim is that the partial derivatives of the objective functional \(E=|1-u_k|\) are thought, optimization, and decision, formed together at the present instant and without iteration. The state satisfies an advection equation whose right-hand side is a dummy source that the structure generates by itself, and substituting \(t=\tau_k\) removes the time variable and lowers the dimension by one. Each new experience adds an increment that decays exponentially with the delay, which we take to be the mechanism of forgetting and of childhood amnesia. In the second half, planning becomes a finite Taylor expansion about the present point, so proving a theorem, writing fiction, and telling a lie are one operation under different objectives. A perfect lie would need infinitely many partial derivatives, and no finite agent can supply them.

Abstract: Generalized Lotk–Volterra (GLV) systems, with roots in ecology, constitute one of the most studied classes of positive ODEs. Recently, a reaction-network perspective for a generalization useful in mathematical epidemiology, called block GLV systems, was offered by Adenane, Avram and Halanay (2026). These authors offer a "siphon calculus" in which the boundary stability of block GLV equations is determined by studying (i) invariant faces associated to minimal siphons, (ii) transversal Jacobians, (iii) invasibility expressed via R-invasion functions, (iv) relay graphs (v) exclusion partitions and (vi) Lyapunov functions, without leaving the original state space. Another reaction-network perspective for GLV systems was offered by Rojas La Luz, Yu and Craciun, who developed a global stability theory for positive equilibria by introducing associated poly-exponential systems obtained through the logarithmic change of variables \(x_i=e^{\xi_i}\). In these logarithmic coordinates, compatibility classes become affine subspaces and remarkably simple quadratic Lyapunov functions establish global convergence of complex-balanced systems. The purpose of the present paper is to combine the two perspectives. We revisit examples considered by Rojas La Luz, Yu and Craciun, where the use of siphon calculus can be avoided, but we still find it useful, at least for suggesting open problems for block GLV systems.

Abstract: Epidemic meningococcal meningitis in the African meningitis belt recurs on two superimposed timescales: a sharp annual dry-season peak and irregular multi-annual epidemics separated by five to twelve years. Because invasive disease is a rare, epidemiologically dead-end outcome of asymptomatic nasopharyngeal carriage, transmission models built on the standard susceptible-infectious template misrepresent the driving process. We formulate a deterministic carriage-structured model in which only carriers transmit, immunity against carriage re-acquisition is leaky, and a conjugate vaccine protects imperfectly and wanes. We show analytically that the basic reproduction number \( R_0 \) is a property of carriage transmission and is decoupled, to first order, from disease incidence, so that reproduction numbers inferred from case notifications estimate the wrong quantity. Using the Castillo-Chavez-Song centre manifold method, we derive, in closed form, the condition for backward bifurcation and prove that it is governed by a threshold \( \varepsilon^\ast \) on carriage-blocking immunity, not by vaccine leakiness or by case-management capacity: the invasive-disease compartment is provably absent from the bifurcation condition. Under seasonal forcing we characterize, through Floquet analysis and two-parameter continuation, the region of immunity-waning and seasonal-amplitude space in which multi-annual recurrence arises, confirming that annual forcing alone cannot generate it. A scenario analysis calibrated to published belt parameter ranges quantifies the burden averted by routine infant immunisation, catch-up campaigns, and improved carriage efficacy, and shows that whether sustained immunisation eliminates epidemics or merely postpones them depends on the same carriage-efficacy threshold that governs bistability.

Abstract: Group-level molecular-interaction evidence is often reduced to pairwise protein-protein interaction graphs before protein-complex detection, potentially discarding experiment membership and over-weighting large groups. Yet prior higher-order studies typically changed the representation, objective, output granularity, and evaluation protocol simultaneously. We isolate the operator-level contribution of higher-order representation by comparing an inverse-size-weighted clique graph, the normalized Zhou hypergraph operator, and a degree-aware penalized hypergraph operator on identical experiment-derived groups. Hyperedges were split by size into training, validation, and test sets; the training split alone defined the protein universe and operators; regularization was selected on validation data; and recovery was evaluated against held-out test hyperedges. Across nine IntAct-derived thematic PSI–MITAB collections and 25 paired initializations per dataset, the penalized hypergraph increased mean held-out symmetric best-match F1 on six datasets after within-dataset Benjamini–Hochberg correction, was indistinguishable on two, and was lower by 0.0015 on Cancer. Positive mean differences ranged from 0.0098 to 0.0356. However, the across-dataset Friedman test did not reach significance (χ² = 5.20, p = 0.074), and the graph–penalized-hypergraph Nemenyi comparison was non-significant (p = 0.111). The penalized operator had the best mean rank (1.39), but secondary metrics and eigengap-selected cluster counts exposed dataset-dependent trade-offs. Observed recovery exceeded degree-preserving and hyperedge-size-preserving nulls on eight of nine datasets. Gene Ontology coherence was high for both representations, with no aspect-level difference surviving multiplicity correction and substantial pooled term overlap (Jaccard 0.78–0.86). Higher-order modeling is therefore conditionally beneficial rather than universally superior: it is most defensible when group membership is retained, regularization is validation-supported, and output granularity is controlled

Abstract: Persistence, coexistence, and boundary transcritical relays are usually studied through model-specific analyses in mathematical epidemiology, ecology, population dynamics, and chemical reaction network theory. Although these fields address closely related questions, they have developed largely independently. This separation is reflected, for example, in the limited mentions of the multi-strain epidemiologic models in ecology’s chemostats and gradostats literature, despite the fact that these are revealed to be very similar once the concept of siphons from chemical reaction network theory is integrated. Conversely, the next-generation matrices and invasion graphs from eco-epidemiology are not mentioned in chemical reaction network theory. Our contribution is firstly conceptual, terminological and definitional: we propose a common framework for the study of boundary phenomena in all positive ODEs subfields. We introduce and formalize notions like reproduction and invasion functions attached to siphon faces, relay graphs, relay tables, and boundary transcritical relays. Some of these concepts are known in one of the above fields but largely absent from the others, while others appear to be new; taken together, they suggest a common language for the analysis of boundary phenomena in positive dynamical systems. The usefulness of the framework is illustrated on multi-strain epidemic models like the Feng-Gavish model, for which we derive explicit, testable conditions. For example, coexistence requires the less fit strain to invade the fitter strain’s equilibrium (for this model, mutual invasibility also ensures coexistence, but explicit further assumptions under which one or the other criterion works for a larger class of models are still unknown). Our approach rests on four pillars: (i) Siphon (a CRN concept) geometry, namely the fact that forward-invariant coordinate faces correspond to siphons, with the disease-free face being the intersection of minimal siphons. (ii) The recently established fact that a transversal Jacobian block on a siphon face is Metzler, which puts under spotlight the roles of its Perron eigenvectors. (iii) A bifurcation theorem linking eigenvalue crossing at a boundary transcritical invasion relay to the emergence of a positive branch on an adjacent face. (iv) Next-generation matrices (NGMs), an MEconcept: on siphon faces, NGMs may be defined via regular splittings, and invasibility may be determined by comparing their spectral radii to > 1.

Abstract: Existing models of the circadian clock in Arabidopsis thaliana are conventionally formulated by integer-order ordinary differential equations (ODEs), which inherently lack the capacity to adequately capture non-local history-dependent regulatory dynamics that arise from sequential biochemical processes such as transcription, translation and protein degradation. Here we construct a Caputo fractional-order model of the Arabidopsis core circadian system and establish local existence and uniqueness, non-negativity, and boundedness of solutions under non-negative initial conditions. Model parameters are fitted to wild-type mRNA expression profiles collected under a standard 12 h light/12 h dark (12L12D) photoperiod. Without any subsequent refitting of parameters, the predictive performance of the fractional-order model is validated on two independent test datasets: wild-type expression time series under three additional photoperiod regimes, and publicly available expression data for major circadian clock loss-of-function mutants. Compared with the original ODE counterpart, the fractional-order formulation exhibits substantially improved performance in reproducing the post-peak decay kinetics and extended tough phase of the PRR5/TOC1 regulatory module. Quantitative error evaluation confirms that the fractional-order model achieves consistently lower mean squared errors across all four photoperiod conditions, and outperforms the integer-order counterpart in two of the four tested mutant backgrounds, indicating that performance gains are not uniformly distributed across all genetic perturbations. Through Matignon-type stability analysis and extensive numerical simulations, we identify a well-defined critical fractional-order threshold. When the fractional order exceeds this critical value, the system’s unique positive equilibrium loses its stability, giving rise to sustained oscillations. From a systems biology perspective, the introduction of fractional-order operators provides a compact phenomenological representation of aggregated historical memory effects and may influence the amplitude, phase, and long-term robustness of the core circadian-clock oscillations. This work offers a new mathematical framework for refining the dynamical characterization of eukaryotic circadian pacemakers.

Abstract: Non-coding RNAs (ncRNAs) play important roles in various biological processes primarily via regulating gene expression at multiple levels. Nevertheless, compared to protein-coding genes, ncRNAs are relatively understudied, characterized by low expression, large variation, and more dynamic expression across different tissues and cell types, which creates unique challenges in differential expression (DE) analysis. We propose a new DE method specifically for ncRNAs based on a composite quantile regression model for count data, in which we combine multiple quantiles and select the best subset of quantiles that most differentiate the expression levels between conditions. The proposed method is flexible, robust and capable of capturing the typically low-count, multi-modal and wide-spread distribution of ncRNA count data. We showed in simulations that our method improved the detection power of ncRNAs while controlling the false positive rates as compared to existing DE tools especially when counts are low. Critical ncRNAs and biologically interpretable targeted pathways were identified when we further applied our method to a Smart-seq-total single-cell RNA-seq dataset and a human organ developmental RNA-seq dataset. An R package to implement CQRM and the codes for this study are publicly available: https://github.com/iamverywell/CQRM.

Abstract: Nowadays, the complexity of electronic health records (EHRs) requires tools capable of efficiently and accurately extracting and interpreting clinically relevant information to support clinicians. This study explores the use of the Cheshire Cat AI framework, configured with Ollama and using LLaMA3 as a language model, with the main purpose of performing automatic analysis of synthetic EHRs from Kaggle. Through specific structured queries, the model was able to successfully reconstruct patients’ clinical histories and extracted useful data such as diagnoses, treatments, visits, comorbidities and demographic data. A validation process through repeated queries was then performed, which confirmed a high level of accuracy. To preserve data privacy, only synthetic datasets were used in this work. Beyond the simple retrieval of information by means of queries, the study highlights the great potential of language models in clinical decision support. Their ability to interpret large and heterogeneous datasets certainly offers new opportunities to improve diagnostic accuracy, simplify workflows and personalise treatments. Specifically, natural language queries by tools such as Cheshire Cat AI can be used for intelligent support systems that can, for instance, integrate multimodal and real-time data to provide medical recommendations. These results represent a first step towards the exploitation of large language models not only for EHR analysis, but also to assist in clinical decision-making processes in different medical fields and, above all, for the study of specific complex diseases such as rare diseases.