Genotype- and Age-Specific Relapse Dynamics in Childhood Acute Lymphoblastic Leukemia: A Longitudinal Mechanistic Study

Genotype- and Age-Specific Relapse Dynamics in Childhood Acute Lymphoblastic Leukemia: A Longitudinal Mechanistic Study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Genotype- and Age-Specific Relapse Dynamics in Childhood Acute Lymphoblastic Leukemia: A Longitudinal Mechanistic Study Wayne Danter This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8653812/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Despite advances in front-line therapy for pediatric acute lymphoblastic leukemia (ALL), relapse remains a leading cause of treatment failure and mortality. Building on prior virtual investigations of TEL-AML1–associated leukemogenesis, this study evaluates post-remission relapse risk across pediatric ALL genotypes using mechanistic simulation. Methods Seven clinically relevant pediatric ALL genotypes were modeled using aiHumanoid v11.0 simulations. Each cohort was simulated from remission (T0) through two post-remission intervals (T1 = 3 months; T2 = 6 months). Longitudinal change was assessed using the Hodges–Lehmann Estimator (HLE), Wilcoxon signed-rank test, and Cliff’s delta. Composite relapse metrics were derived from 12 relapse-related outcome domains. Results Early relapse signals were detected in high-risk genotypes, including TEL-AML1 + PAX5 deletion, TEL-AML1 + radiation exposure, and TEL-AML1 + Ikaros deletion, characterized by elevated HLE values and large effect sizes. Low-risk genotypes, including TEL-AML1 alone and TEL-AML1 + family history, showed minimal deviation from remission through 6 months. Conclusions HLE-based longitudinal analysis enables detection of genotype-specific relapse dynamics in pediatric ALL and supports individualized, time-aware post-remission surveillance strategies. Health sciences/Medical research/Preclinical research Biological sciences/Biotechnology/Genomics/Pharmacogenomics Acute Lymphoblastic Leukemia Pediatric Oncology aiHumanoid Hodges–Lehmann Estimator Genetic Relapse Risk Simulation Study TEL-AML1 PAX5 Radiation Figures Figure 1 Figure 2 INTRODUCTION Acute lymphoblastic leukemia (ALL) is the most common childhood malignancy, accounting for approximately 25% of pediatric cancers [ 3 ]. Contemporary treatment regimens have yielded remission rates exceeding 90% [ 4 ], yet relapse remains the leading cause of ALL-related mortality [ 5 , 6 , 26 ]. Disease recurrence is often insidious, emerging without clear clinical warning during post-remission surveillance. Identifying which patients are at increased risk of relapse, and when that relapse may occur, is essential for improving long-term survival and guiding follow-up care [ 8 ][ 9 ]. Although minimal residual disease (MRD) assessment has become central to current ALL risk stratification, it is an imperfect predictor of relapse [ 22 , 24 ]. MRD measurements do capture residual leukemic burden at discrete timepoints but do not fully reflect underlying biological processes including clonal evolution, immune evasion, or treatment resistance that may develop during post-remission surveillance [ 23 ]. As a result, patients with initially favorable MRD profiles may still experience relapse, while others undergo prolonged monitoring without clear benefit. These limitations underscore the need for complementary approaches that capture relapse course rather than static burden alone. In our previously published study, “Investigating Genetic and Familial Risks in Childhood ALL: A Longitudinal Virtual Study Using aiHumanoid [ 12 ] Simulations of TEL-AML1 Gene Fusion”, we used the aiHumanoid [ 12 ] platform to explore early leukemogenic effects of the TEL-AML1 fusion, including its modulation by family history (FHx). That work demonstrated that virtual modeling could recapitulate clinically meaningful early disease behavior, including precursor immune dysfunction and hematologic changes. Building on those results, the present study addresses an important clinical question: how does relapse risk evolve over time in children with different molecular subtypes of ALL, once remission has been achieved? To answer this, we conducted longitudinal simulations of seven common pediatric ALL genotypes—each known to have distinct prognostic implications. These simulations tracked 12 outcome domains related to relapse progression from the time of remission (T0) through two clinically relevant time points: T1 (3 months) and T2 (6 months). Increasing evidence suggests that relapse risk in pediatric ALL is developmentally modulated, reflecting age-dependent interactions among immune reconstitution, bone marrow microenvironmental dynamics, and therapy responsiveness. Consequently, relapse vulnerability may emerge preferentially within specific developmental windows rather than remaining constant over time. Incorporating age as an active biological dimension is therefore essential for accurate post-remission risk stratification. To detect subtle but important changes in relapse risk, we employed a robust nonparametric statistical framework incorporating the Wilcoxon signed-rank test, and Cliff’s delta [ 20 ] effect size estimates, and the Hodges–Lehmann [ 21 ] Estimator (HLE). This triad enables high-resolution detection of both population-level and subject-level trends in disease behavior. We further developed composite relapse indices to support rapid, quantitative risk classification for each genotype. This study represents a next-generation AI/ML application of physiologically grounded simulation to anticipate relapse and personalize surveillance in childhood ALL. METHODS This study used the aiHumanoid [ 12 ] v11 simulation system, a high-fidelity fuzzy cognitive map FCM)-based platform designed to model virtual pediatric physiology and disease progression. Seven of the more common pediatric ALL genotypes (mutations) were selected based on known clinical significance and included: TEL-AML1, TEL-AML1 + FHx, TEL-AML1 + Ikaros Del, TEL-AML1 + PAX5 Del, TEL-AML1 + Radiation exposure, TEL-AML1 + TEL LOH, and TEL-AML1 + p16 Del. Each genotype cohort consisted of 25 virtual subjects (n = 175 total), aged 1–20 years (6 cohorts), modeled from the point of achieving remission (T0). Subjects were followed through two post-remission checkpoints: T1 (3 months) and T2 (6 months). All simulation inputs were standardized using the − 1 to + 1 scale for FCM inputs. To quantify relapse dynamics within the simulation, two composite metrics were implemented as standalone fuzzy cognitive map (FCM) nodes: the ALL_Relapse Risk Score and the ALL_Relapse Pressure Index (RPI). These metrics respectively estimate snapshot relapse probability and trajectory-based relapse momentum, drawing from weighted input domains related to clonal evolution, minimal residual disease, treatment resistance [ 7 ][ 10 ][ 11 ], and immune evasion. A detailed breakdown of input structure, functional behavior, and operational calculation for both metrics is provided in Appendix 2. Table 1: The twelve outcome domains were monitored at each timepoint, include: Immune Activation → The stimulation of immune cells to recognize and respond to threats, including leukemic cells. ALL_Clonal Evolution → The stepwise genetic diversification of leukemic cells, leading to more aggressive or therapy-resistant subclones. ALL_Development → The biological process through which normal progenitor cells transform into malignant lymphoblasts in acute lymphoblastic leukemia [3][4]. ALL_Facilitating the Immune Response to ALL → Mechanisms or interventions that enhance immune detection and destruction of ALL cells. ALL_Inhibiting the Immune Response to ALL → Processes by which leukemic cells suppress or evade immune surveillance, promoting disease persistence or relapse. ALL_Minimal Residual Disease (BC-ALL) → The small number of leukemic cells that remain after treatment in B-cell ALL, undetectable by standard microscopy but detectable by sensitive molecular methods. ALL_Pre-leukemic Clone Expansion → The growth of early abnormal cell populations with leukemic potential prior to full malignant transformation. ALL_Relapse → The reappearance or progression of leukemia after a period of remission, typically due to treatment-resistant clones. ALL_Relapse Pressure Index → A composite measure quantifying the overall biological momentum or stress toward relapse based on longitudinal outcome trends. ALL_Relapse Risk Score → A summary metric estimating the likelihood of relapse for a given subject or genotype, derived from key indicators like HLE, MRD, and clonal shifts. ALL_Treatment Resistance → The failure of leukemic cells to respond to chemotherapy or immunotherapy, often due to genetic mutations or protective microenvironments. QoL-Physical wellbeing (PW) → A patient’s physical health status and ability to perform daily activities, serving as a proxy for treatment burden and recovery. Statistical evaluation used a three-part nonparametric approach: Wilcoxon Signed-Rank Test: Assessed whether the paired differences significantly deviated from zero. The conservative Bonferroni correction was selected to prioritize specificity over sensitivity given the clinical implications of false-positive relapse signals. The corrected p value for statistical significance is P<0.004. Cliff’s Delta: Quantified the effect size of change, providing population-level directionality and strength of the shift. A Cliff’s delta threshold of |δ| ≥ 0.474 corresponds to a large effect, chosen to minimize false-positive relapse signals in a high-dimensional longitudinal virtual study. Hodges–Lehmann [21] Estimator (HLE) for paired data: Calculated as the median of all pairwise differences between paired timepoints (T1-T0, T2-T0, and T2-T1) for each outcome and subject, estimating the typical per-subject change. Composite relapse metrics were derived: - ALL_Relapse Score: - ALL_Relapse Pressure Index: All relapse ‘risk’ metrics in this study represent relative biological propensity within the simulated system, not absolute clinical probability. All results were normalized to the system-standard -1 to +1 scale, and SST (Sober Second Thought) methodology was used for verification. While the aiHumanoid simulation platform is proprietary, all model logic relevant to this study—including outcome domain definitions, composite metric construction, statistical thresholds, and interpretive criteria—has been fully specified to support independent conceptual replication and methodological scrutiny without requiring access to source code. RESULTS T1 vs T0: Early Recurrence Risk Following Remission in Pediatric Virtual ALL This longitudinal in silico study evaluated early recurrence patterns in pediatric acute lymphoblastic leukemia (ALL) [ 3 , 4 ] across 175 virtual subjects, each stratified by one of seven TEL-AML1–based genotypic profiles and six developmental age cohorts (0, 2–3, 5, 10, 15, and 20 years). Subjects were evaluated across 12 relapse-relevant outcome domains at two time points: T0 (remission) and T1 (3 months post-remission). The Hodges–Lehmann [ 21 ] Estimator (HLE) was used to estimate subject-level shifts, Cliff’s delta [ 20 ] quantified population-level effect sizes, and the Wilcoxon signed-rank test assessed statistical significance. A Risk Flag was applied when p ± 0.1 or |Cliff’s delta [ 20 ]| ≥ 0.474. Figure 1 summarizes the distribution of HLE values across domains and age cohorts for the T1 vs T0 comparison. Genotype-Linked Patterns of Early Relapse Among the seven genotypes modeled, TEL-AML1 + PAX5 Deletion and TEL-AML1 + Radiation Exposure emerged as the highest risk groups for early recurrence post-remission. TEL-AML1 + PAX5 Del: Demonstrated 369 flagged domain events across all age strata. At age 10, 9 of 12 domains were flagged, including Clonal Evolution (HLE = + 0.22, Cliff’s d = 0.54, p = 0.004) and Minimal Residual Disease (MRD) (HLE = + 0.18, d = 0.49, p = 0.003). This genotype consistently activated early relapse pathways from age 5 onward. TEL-AML1 + Radiation Exposure: Produced the highest total burden with 379 flagged events. At age 15, Treatment Resistance (HLE = + 0.19, d = 0.51, p = 0.002) and Clonal Evolution showed strong relapse indicators. This genotype exhibited sustained relapse pressure from mid-childhood to early adulthood. Intermediate-risk genotypes such as TEL-AML1 + p16 Deletion and TEL-AML1 + Ikaros Deletion showed limited and age-dependent relapse activity. In contrast, TEL-AML1 alone and TEL-AML1 + FHx (family history) showed minimal biological deviation from remission, with stable Wilcoxon p-values (p > 0.004) and consistently low HLE and Cliff’s d values across all ages. T2 vs T0: Early Relapse Risk at 6 Months Post-Remission in Pediatric ALL To investigate the potential for early subclinical relapse, we analyzed virtual longitudinal profiles comparing T2 (6 months post-remission) with T0 (remission baseline) across pediatric acute lymphoblastic leukemia [ 3 ][ 4 ] (ALL) subjects. This comparison was conducted using three complementary statistical approaches: the Hodges–Lehmann [ 21 ] Estimator (HLE) for subject-level shifts, Cliff’s delta [ 20 ] for population-level effect sizes, and the Wilcoxon signed-rank test for statistical significance. Figure 2 presents the HLE estimates for T2 vs T0 across the 12 relapse domains and six age cohorts. We defined high-risk relapse strata as those that met the following criteria: Wilcoxon p 0.1, and |Cliff’s delta [ 20 ]| ≥ 0.474. This composite model identifies both clinically relevant and statistically robust relapse signals. Out of 72 possible age × outcome combinations (6 age cohorts × 12 outcome domains), 18 were flagged as high risk (25%). The most pronounced risk signals were seen in the 2-3-year and 5-year age cohorts, which together accounted for 11 of the 18 flagged combinations. Age-Dependent Patterns: The 2-3-year and 5-year cohorts showed the highest burden of relapse risk, with major contributions from clonal evolution, minimal residual disease (MRD), relapse pressure index, and treatment resistance [ 7 ][ 10 ][ 11 ]. The 10-year group also showed moderate relapse activity, while ages 0 and 20 were relatively stable. Domain-Specific Insights: Among the 12 relapse-related domains, four were especially prominent in driving high-risk profiles: ALL_Clonal Evolution: Frequently flagged across ages 2–3 to 10 with HLE > 0.14 and Cliff’s delta [ 20 ] > 0.50. ALL_Minimal Residual Disease (BC-ALL): Persistent positivity in age 5 cohort (HLE = + 0.230, Cliff’s d = 0.51, Wilcoxon p = 0.002). ALL_Treatment Resistance: Flagged in ages 5–15, suggesting reemergence of drug-resistant subclones. ALL_Relapse Pressure Index: Increased systemic stress in leukemic niches was detected in ages 2–3–10. Illustrative High-Risk Stratum: At age 5, Minimal Residual Disease (MRD) showed a subject-level increase (HLE = + 0.230), a large effect size (Cliff’s d = 0.51), and strong statistical support (Wilcoxon p = 0.002), fulfilling all three criteria for high-risk flagging. T2 vs T1: Late Subclinical Recurrence Risk at 6 Months vs 3 Months Post-Remission To assess continued relapse risk trajectories after initial post-remission surveillance, we compared virtual pediatric ALL profiles between T2 (6 months post-remission) and T1 (3 months post-remission) across 175 subjects. Using Hodges–Lehmann [ 21 ] Estimator (HLE), Cliff’s delta [ 20 ], and Wilcoxon signed-rank test, we identified relapse-relevant changes at the subject level, population level, and statistical level, respectively. We defined high-risk strata as those with Wilcoxon p 0.1, and |Cliff’s delta [ 20 ]| ≥ 0.474. Out of 72 age × domain strata, 21 were flagged as high risk (29.2%), indicating an increased burden of biological recurrence activity in this interval compared to earlier comparisons. The 2-3-year, 5-year, and 10-year age cohorts were especially vulnerable, contributing 16 of the 21 high-risk strata (76%). Age-Stratified Patterns 2–3 Years: Clonal Evolution, MRD, and Treatment Resistance continued to show increasing divergence (e.g., MRD HLE = + 0.19, Cliff’s d = 0.50, p = 0.002). 5 Years: Relapse Pressure Index and Resistance metrics showed strong relapse signaling. 10 Years: A resurgence of Clonal Evolution and Resistance domains reflected cumulative relapse burden. Older ages (15, 20) exhibited partial stabilization, with only 2 domains flagged in total, suggesting possible biological resolution or treatment buffering. Domain-Specific Insights Three domains remained most predictive of recurrence risk in this later window: ALL_Clonal Evolution: Persistent across 2–3 to 10 years, with several strata exceeding Cliff’s d = 0.50 and HLE > + 0.15. ALL_Minimal Residual Disease (BC-ALL): Especially in 2–3 and 5 years, indicating persistent leukemic cell populations. ALL_Treatment Resistance: Suggests evolving therapy escape mechanisms in intermediate ages. One notable addition was subtle emergence of ALL_Relapse Risk Score in older children (age 10–15), with HLE > + 0.11 and Cliff’s d > 0.45, though not reaching all criteria thresholds (p > 0.004). High-Risk Stratum Example At age 2–3, MRD was again prominently elevated (HLE = + 0.19, Cliff’s d = 0.50, Wilcoxon p = 0.002), meeting all flagging thresholds and reinforcing its role as a core recurrence driver. Summary: Risk Stratification of Early and Subclinical Recurrence in Pediatric ALL Analysis across three post-remission intervals (T0 vs T1, T0 vs T2, and T2 vs T1) identified consistent genotype- and age-specific patterns of relapse vulnerability [ 5 ][ 6 ] in pediatric acute lymphoblastic leukemia [ 3 ][ 4 ] (ALL). Subjects were classified into high, intermediate, or low risk categories based on the number of flagged relapse domains, strength of statistical signals (p 0.1, |Cliff’s d| ≥ 0.474), and persistence across time points. High-Risk Genotypes These subtypes showed recurrent, multi-domain relapse activation across all three-time intervals, particularly during mid-childhood: TEL-AML1 + PAX5 Deletion Persistent MRD elevation, Clonal Evolution, and Treatment Resistance from age 5 onward. Up to 9 of 12 relapse domains flagged at age 10. Consistent early and late recurrence indicators. TEL-AML1 + Radiation Exposure Highest total number of flagged domains. Sustained relapse signaling in Treatment Resistance and Clonal Evolution from ages 10 to 15. Strong relapse pressure despite initial remission. Recommended action: Intensified post-remission surveillance, especially between ages 5–15. Intermediate-Risk Genotypes These groups showed limited or age-dependent recurrence features, often confined to one time point or fewer domains: TEL-AML1 + p16 Deletion TEL-AML1 + Ikaros Deletion Demonstrated relapse signatures in selected age strata (typically 5–10 years). Partial but non-persistent elevation of MRD or Clonal Evolution. Recommended action: Age-specific follow-up, especially during mid-childhood. Low-Risk Genotypes These profiles remained biologically stable across all time points with negligible domain flagging: TEL-AML1 Alone (Baseline) TEL-AML1 + FHx (Family History) No significant domain activation (Wilcoxon p > 0.004 across all comparisons). Consistently low HLE and Cliff’s delta [ 20 ] metrics. Recommended action: Standard surveillance schedules appear sufficient. Age-Based Recurrence Risk Patterns Relapse vulnerability exhibited a developmentally stratified pattern that interacted with genotype-specific risk: • Highest Risk Ages: Ages 2–3, 5, and 10 years demonstrated the greatest burden of flagged relapse domains—particularly MRD, Clonal Evolution, and Treatment Resistance—in high-risk genotypes (e.g., PAX5 Del, Radiation). • Lowest Risk Ages: Infants (age 0) and older adolescents (age 20) consistently showed minimal relapse signal across genotypes, suggesting relative resilience or transient therapeutic sensitivity. This developmental gradient reinforces the need for age-stratified surveillance, particularly during mid-childhood in genetically high-risk patients. DISCUSSION This study expands upon our previously published virtual modeling framework for childhood ALL by introducing a longitudinal, genotype- and age-stratified analysis of post-remission relapse risk. Using aiHumanoid [ 12 ] v11.0, we simulated disease trajectories across 175 virtual pediatric subjects, incorporating seven clinically relevant TEL-AML1-based genotypes and six developmental age cohorts. Outcomes were evaluated across 12 biologically grounded relapse domains at three key time points (T0, T1, T2). The central innovation of this work lies in the combined use of the Hodges–Lehmann [ 21 ] Estimator (HLE), Cliff’s delta [ 20 ], and the Wilcoxon signed-rank test. Together, these metrics enabled high-resolution detection of both subject-level and population-level shifts in relapse risk. HLE proved valuable in quantifying early subclinical changes that may precede overt clinical relapse. Risk Stratification by Genotype and Age A key insight from this analysis is the emergence of distinct risk tiers based on genotype: High-risk genotypes (TEL-AML1 + PAX5 Deletion and TEL-AML1 + Radiation Exposure) show consistent, multi-domain relapse activation across all time points, with up to 9 of 12 outcome domains flagged in mid-childhood cohorts. These genotypes represent urgent candidates for intensified post-remission surveillance and potentially early intervention trials. Intermediate-risk genotypes (TEL-AML1 + Ikaros Deletion and TEL-AML1 + p16 Deletion) demonstrated transient or age-restricted relapse signals, particularly between ages 5–10. While less aggressive than high-risk profiles, they may still benefit from tailored monitoring strategies during key developmental windows. Low-risk genotypes (TEL-AML1 alone and TEL-AML1 + FHx) remained biologically stable throughout, with negligible HLE shifts and consistently non-significant Wilcoxon p-values. These subgroups may be candidates for reduced-intensity follow-up, limiting unnecessary exposure to monitoring-associated burdens. Developmental Windows of Vulnerability Importantly, relapse risk also demonstrated a developmentally stratified pattern, with the ages 2–3, 5, and 10 years consistently associated with the highest burden of flagged relapse domains, independent of genotype. These ages likely reflect critical intersections of immune reconstitution, bone marrow microenvironmental dynamics, and therapy responsiveness. In contrast, infants (age 0) and older adolescents (age 20) consistently showed biological stability, suggesting temporary windows of resilience. This dual-axis stratification (by genotype and age) provides a powerful framework for anticipating relapse dynamics and refining post-remission care. The visual trends shown in Figs. 1 and 2 reinforce the emergence of MRD, Clonal Evolution, and Treatment Resistance as dominant relapse drivers [ 8 ][ 9 ][ 10 ] during mid-childhood. In contrast to most “digital twin” implementations that rely primarily on static predictive models or correlative machine learning [14,15,25], the aiHumanoid framework employed here is explicitly causal, physiologically grounded, and dynamically adaptive. The system models disease progression as an evolving network of interacting biological processes, enabling inference about relapse momentum, temporal escalation, and counterfactual trajectories rather than point estimates alone. This distinction is particularly important in post-remission surveillance, where early biological drift may precede measurable clinical relapse. By capturing trajectory-level dynamics across genotype and developmental stage, this approach extends beyond risk scoring toward anticipatory disease modeling. Simulation Metrics and Clinical Relevance The derived composite indices— ALL_Relapse Score and Relapse Pressure Index (RPI) —proved effective in summarizing relapse dynamics. When paired with primary markers such as MRD and Clonal Evolution, these metrics enable a structured approach to early warning detection. With further validation, they could form the basis of real-time clinical dashboards powered by patient-specific genomic and treatment data. Limitations and Future Directions This work is grounded in high-fidelity, physiology-informed virtual simulations. However, these findings should be interpreted with caution until validated against real-world longitudinal datasets. Future work will focus on calibrating simulation-derived risk metrics with clinical relapse outcomes and extending the model to include adaptive therapies or immune-targeted strategies. As longitudinal pediatric relapse datasets mature and federated data-sharing initiatives expand, prospective calibration of simulation-derived relapse metrics against real-world outcomes will become increasingly feasible. Reproducibility Statement While the aiHumanoid simulation engine is proprietary, all causal structures, outcome domains, composite metric definitions, thresholds, and statistical criteria required for independent conceptual replication are fully specified in this manuscript. Independent validation can be performed using alternative causal simulation frameworks implementing equivalent domain logic. Conclusion This virtual longitudinal study demonstrates the potential feasibility and utility of combining HLE, Cliff’s delta [ 20 ], and Wilcoxon testing to quantify genotype-specific relapse risk in childhood ALL. The results identify high-risk and low-risk genotypes early in the surveillance period, offering a pathway for precision follow-up care. Our approach supports individualized post-remission planning, provides deeper insights into the biological underpinnings of relapse, and establishes a scalable method for evaluating new genomic variants. The integration of advanced statistical tools and simulation modeling offers a powerful paradigm for anticipating and managing relapse in pediatric oncology. Declarations Competing interests Wayne R. Danter is the Founder and developer of the aiHumanoid technology and serves as Chief Scientific Officer of HumanQAI, Inc. (formerly 123Genetix, Inc.). All other authors declare no competing interests. Funding This research received no external funding. Ethics approval and consent to participate Not applicable. This study was conducted entirely in silico using population-based virtual pediatric acute lymphoblastic leukemia models. No human participants, patient-level clinical data, or identifiable biological specimens were involved; therefore, ethics approval and informed consent were not required. Consent for publication Not applicable. No human participants or identifiable data were included in this study. Data availability The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request. Data are not publicly available due to the proprietary nature of the aiHumanoid simulation platform. Author contributions Conceptualization: W.R.D.; Methodology: W.R.D.; Formal analysis: W.R.D.; Writing—original draft: W.R.D.; Writing—review and editing: all authors. Acknowledgements The authors acknowledge limited use of GPT-5.1 (OpenAI) as a supporting tool in the development and refinement of population-based virtual subject representations within the aiHumanoid simulation framework. GPT-5.1 was not used to generate real patient data, to perform statistical analyses, or to make independent scientific or clinical decisions. All model design choices, simulation parameters, data interpretation, and conclusions were defined, reviewed, and approved by the authors, who retain full responsibility for the work. Clinical trial registration Not applicable. References Farber S, Diamond LK, Mercer RD, Sylvester RF, Wolff JA. Temporary remissions in acute leukemia in children produced by folic acid antagonist, 4-aminopteroyl-glutamic acid (aminopterin). N Engl J Med. 1948;238(23):787–93. Pinkel D, Simone JV, Hustu HO, Aur RJ. Total therapy of acute lymphocytic leukemia. Pediatrics. 1971;48(5):846–57. Inaba H, Greaves M, Mullighan CG. Acute lymphoblastic leukaemia. Lancet . 2013;381(9881):1943–55. Pui CH, Yang JJ, Hunger SP, Pieters R, Schrappe M, Biondi A, et al. Childhood acute lymphoblastic leukemia: progress through collaboration. J Clin Oncol. 2015;33(27):2938–48. Oskarsson T, Söderhäll S, Arvidson J, Forestier E, Montgomery S, Frandsen TL, et al. Relapsed childhood acute lymphoblastic leukaemia in the Nordic countries: risk factors, treatment and outcome. Haematologica. 2016;101(1):68–76. Ceppi F, Cazzaniga G, Colombini A, Biondi A, Conter V, et al. Risk factors for relapse in childhood acute lymphoblastic leukemia. Leuk Lymphoma. 2015;56(6):1907–18. Hughes A, Wade R, Richards S, Hancock J, Vora A, Mitchell C, et al. Treatment and outcome of relapsed childhood T-cell acute lymphoblastic leukaemia: results from the UKALLR3 trial. Haematologica. 2020;105(3):574–82. Rheingold SR, Salzer WL, Devidas M, Borowitz MJ, Carroll AJ, Gastier-Foster JM, et al. Determinants of survival after first relapse of acute lymphoblastic leukemia. Leukemia. 2024;38(2):456–68. Del Rocco NJ, Annesley CE, Alexander TB, Bartram J, Attarbaschi A, Moorman AV, et al. Enhanced risk stratification for children and young adults with ALL: the PI-UKALL score. Leukemia. 2024;38(6):1231–42. Eckert C, von Stackelberg A, Seeger K, Groeneveld-Krentz S, Cario G, Schrauder A, et al. Risk factors and outcomes in children with high-risk B-cell acute lymphoblastic leukemia. Cancer Lett. 2021;499:154–64. Slayton WB, Devidas M, Hunger SP, Heerema NA, Carroll AJ, Winick NJ, et al. Poverty and relapse risk in children with acute lymphoblastic leukaemia. Blood. 2023;142(3):221–32. Danter WR. Investigating genetic and familial risks in childhood ALL: a longitudinal virtual study using aiHumanoid simulations of TEL-AML1 gene fusion. medRxiv . 2024. doi: 10.1101/2024.06.19.24309162 . Katsoulakis E, Cloutier M, Vasan A, Bonnet A, Lemaire J, Shah NR, et al. Digital twins for health: a scoping review. NPJ Digit Med. 2024;7:110. Sadée C, Brochard J, van der Garde M, Chouvarda I, Garcia-Remesal M, Roca J, et al. Medical digital twins: enabling precision medicine and personalized healthcare delivery. Lancet Digit Health. 2025;7(2):e123-34. Giansanti D, Simeoni R, Barcellona A, Mercuri NB, Morabito FC, et al. Exploring the potential of digital twins in cancer treatment. J Clin Med. 2025;14(10):3574. Zhuang Y, Zhao J, Li H, Zhang T, Sun H, et al. Advancing precision oncology with digital and virtual twins. Front Oncol. 2025;15:119–32. Li X, Zhou Y, Chen Q, Wang J, Liu S, et al. From virtual to reality: innovative practices of digital twins in tumor therapy. J Transl Med. 2025;23:117. Demir A, Chen Y, Gómez J, Zhou H, Lee J, et al. Towards verifiable cancer digital twins: tissue-level modelling and beyond. Front Physiol. 2024;15:1473125. Bruynseels K, Santoni de Sio F, van den Hoven J. Exploring approaches for predictive cancer patient digital twins. Front Oncol. 2020;10:9586248. Cliff N. Dominance statistics: ordinal analyses to answer ordinal questions. Psychol Bull. 1993;114(3):494–509. Hodges JL, Lehmann EL. Estimates of location based on rank tests. Ann Math Stat. 1963;34(2):598–611. Kim, H., Yoon, S.H., Kang, S. et al. Improved survival in pediatric acute lymphoblastic leukemia through therapy intensification based on minimal residual disease and protocol-driven early response risk classification. Blood Res. 60, 40 (2025). https://doi.org/10.1007/s44313-025-00085-3 Ștefan A-I, Radu L-E, Jardan D and Coliță A (2025) The emerging role of next-generation sequencing in minimal residual disease assessment in acute lymphoblastic leukemia: a systematic review of current literature. Front. Med . 12:1570041. doi: 10.3389/fmed.2025.1570041 Menon H, Singh PK, Bagal B, Dolai T, Jain A, Chaudhri A. Minimal Residual Disease in the Management of B-Cell Acute Lymphoblastic Leukemia: A Systematic Review of Studies from Indian Settings. Indian J Hematol Blood Transfus. 2024;40(1):1–11. doi: 10.1007/s12288-023-01641-6. Vallée A. Digital Twins for Personalized Medicine Require Epidemiological Data and Mathematical Modeling: Viewpoint J Med Internet Res 2025;27:e72411 doi: 10.2196/72411 NCI PDQ (2025)-PDQ® Pediatric Treatment Editorial Board. PDQ Childhood Acute Lymphoblastic Leukemia Treatment. Bethesda, MD: National Cancer Institute. 2025, Available at: https://www.cancer.gov/types/leukemia/hp/child-all-treatment-pdq . Additional Declarations Yes there is potential conflict of interest. Supplementary Files Appendix13.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8653812","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":578225104,"identity":"19c7a364-fcc9-476b-8a1a-374faf06fa7e","order_by":0,"name":"Wayne Danter","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAwElEQVRIiWNgGAWjYFCCBDApByIOPCBaywEGBmOwlgRStCQ2IGwkAAyOJz97/KHiXvr8sMMPgbbYyek2ENJy5pm5wYEzxbkbb6cZALUkG5sdIKDF7EaCmcTBtoTcjbMTQFoOJG4jrCX9m8TBfwnphrPTPxCrJQdoS0NCgrx0DpG22J95UyZx5liC4QbpnIIDCQZE+EWyPX2bREVNgrz87PTNHz5U2MkR1AIHBmCVBsQqBwH5BlJUj4JRMApGwYgCAMYgS1TalmGDAAAAAElFTkSuQmCC","orcid":"","institution":"HumanQAI","correspondingAuthor":true,"prefix":"","firstName":"Wayne","middleName":"","lastName":"Danter","suffix":""}],"badges":[],"createdAt":"2026-01-21 00:56:25","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8653812/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8653812/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":101075025,"identity":"a56089e5-b33b-4502-a267-9e49e2bd8d04","added_by":"auto","created_at":"2026-01-25 10:24:45","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":149299,"visible":true,"origin":"","legend":"","description":"","filename":"ExpandedALLRecurrenceManuscriptCompleteV21.docx","url":"https://assets-eu.researchsquare.com/files/rs-8653812/v1/9dbca5bb379a6be468e77280.docx"},{"id":101075246,"identity":"e854b8df-aedf-4b2a-b73d-f807c3828802","added_by":"auto","created_at":"2026-01-25 10:25:16","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3278,"visible":true,"origin":"","legend":"","description":"","filename":"26LEU0122.json","url":"https://assets-eu.researchsquare.com/files/rs-8653812/v1/97f21ce0062a74c58e50943c.json"},{"id":101075181,"identity":"674aa5da-8199-404f-adea-c5df9837ed24","added_by":"auto","created_at":"2026-01-25 10:25:09","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":88377,"visible":true,"origin":"","legend":"","description":"","filename":"26LEU01220enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8653812/v1/ddaaf9c0d3ee37f0c92e56f8.xml"},{"id":101075004,"identity":"19363c71-b1fe-4dc3-9981-fda3f64436b2","added_by":"auto","created_at":"2026-01-25 10:24:36","extension":"jpeg","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2544881,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8653812/v1/aecafadbbf668cebd93309cd.jpeg"},{"id":101075126,"identity":"9d42c9cb-3fc9-47fa-93c1-e80094cb6fd5","added_by":"auto","created_at":"2026-01-25 10:25:01","extension":"emf","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":790636,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.emf","url":"https://assets-eu.researchsquare.com/files/rs-8653812/v1/25d56a6a8336a86420138219.emf"},{"id":101075027,"identity":"72e49e64-e588-46f7-905f-c2fd92d236b9","added_by":"auto","created_at":"2026-01-25 10:24:45","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":437985,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8653812/v1/83cee43f2006407e7866dc96.png"},{"id":101075059,"identity":"ae24e3f0-642c-4d5c-aad4-72edd2c5d5e6","added_by":"auto","created_at":"2026-01-25 10:24:49","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":157163,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8653812/v1/77e257682324a8854de6aa91.png"},{"id":101075245,"identity":"8b7ea758-56ea-4f91-8558-004c3e160125","added_by":"auto","created_at":"2026-01-25 10:25:16","extension":"xml","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":86013,"visible":true,"origin":"","legend":"","description":"","filename":"26LEU01220structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8653812/v1/82fbd21ac636f9a9f78ebf6c.xml"},{"id":101075020,"identity":"24b4c832-1386-42b2-9a71-7625645f9b21","added_by":"auto","created_at":"2026-01-25 10:24:43","extension":"html","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":99292,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8653812/v1/0ba6f6290657c7804240c978.html"},{"id":101075029,"identity":"530eb68f-8028-4600-ac1e-a623c0eef3d3","added_by":"auto","created_at":"2026-01-25 10:24:46","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":318384,"visible":true,"origin":"","legend":"\u003cp\u003eHLE estimates for Time 1 (3 months) vs Time 0\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8653812/v1/8c11b6be61cdebe2d9778d41.png"},{"id":101075034,"identity":"f86fe23e-ec4a-4f1e-aada-afdbc7966599","added_by":"auto","created_at":"2026-01-25 10:24:47","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":278943,"visible":true,"origin":"","legend":"\u003cp\u003eHLE estimates for Time 1 (6 months) vs Time 0\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8653812/v1/10be03901a682652fd9b9c4a.png"},{"id":102295137,"identity":"1f4b7008-0746-4a83-a4e3-9ccd6bf1b031","added_by":"auto","created_at":"2026-02-10 10:09:09","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":784809,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8653812/v1/5381a2ce-19e8-4c95-b389-562662930f65.pdf"},{"id":101075055,"identity":"52499ed7-97ac-47e5-bf83-7eea97a86887","added_by":"auto","created_at":"2026-01-25 10:24:49","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":18915,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix13.docx","url":"https://assets-eu.researchsquare.com/files/rs-8653812/v1/7ac0b9f07e766ad0ef502bf3.docx"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential conflict of interest.","formattedTitle":"Genotype- and Age-Specific Relapse Dynamics in Childhood Acute Lymphoblastic Leukemia: A Longitudinal Mechanistic Study","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eAcute lymphoblastic leukemia (ALL) is the most common childhood malignancy, accounting for approximately 25% of pediatric cancers [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Contemporary treatment regimens have yielded remission rates exceeding 90% [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e], yet relapse remains the leading cause of ALL-related mortality [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. Disease recurrence is often insidious, emerging without clear clinical warning during post-remission surveillance. Identifying which patients are at increased risk of relapse, and when that relapse may occur, is essential for improving long-term survival and guiding follow-up care [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e][\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAlthough minimal residual disease (MRD) assessment has become central to current ALL risk stratification, it is an imperfect predictor of relapse [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. MRD measurements do capture residual leukemic burden at discrete timepoints but do not fully reflect underlying biological processes including clonal evolution, immune evasion, or treatment resistance that may develop during post-remission surveillance [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. As a result, patients with initially favorable MRD profiles may still experience relapse, while others undergo prolonged monitoring without clear benefit. These limitations underscore the need for complementary approaches that capture relapse course rather than static burden alone.\u003c/p\u003e \u003cp\u003eIn our previously published study, \u0026ldquo;Investigating Genetic and Familial Risks in Childhood ALL: A Longitudinal Virtual Study Using aiHumanoid [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] Simulations of TEL-AML1 Gene Fusion\u0026rdquo;, we used the aiHumanoid [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] platform to explore early leukemogenic effects of the TEL-AML1 fusion, including its modulation by family history (FHx). That work demonstrated that virtual modeling could recapitulate clinically meaningful early disease behavior, including precursor immune dysfunction and hematologic changes.\u003c/p\u003e \u003cp\u003eBuilding on those results, the present study addresses an important clinical question: how does relapse risk evolve over time in children with different molecular subtypes of ALL, once remission has been achieved? To answer this, we conducted longitudinal simulations of seven common pediatric ALL genotypes\u0026mdash;each known to have distinct prognostic implications. These simulations tracked 12 outcome domains related to relapse progression from the time of remission (T0) through two clinically relevant time points: T1 (3 months) and T2 (6 months).\u003c/p\u003e \u003cp\u003eIncreasing evidence suggests that relapse risk in pediatric ALL is developmentally modulated, reflecting age-dependent interactions among immune reconstitution, bone marrow microenvironmental dynamics, and therapy responsiveness. Consequently, relapse vulnerability may emerge preferentially within specific developmental windows rather than remaining constant over time. Incorporating age as an active biological dimension is therefore essential for accurate post-remission risk stratification.\u003c/p\u003e \u003cp\u003eTo detect subtle but important changes in relapse risk, we employed a robust nonparametric statistical framework incorporating the Wilcoxon signed-rank test, and Cliff\u0026rsquo;s delta [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] effect size estimates, and the Hodges\u0026ndash;Lehmann [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] Estimator (HLE). This triad enables high-resolution detection of both population-level and subject-level trends in disease behavior. We further developed composite relapse indices to support rapid, quantitative risk classification for each genotype.\u003c/p\u003e \u003cp\u003eThis study represents a next-generation AI/ML application of physiologically grounded simulation to anticipate relapse and personalize surveillance in childhood ALL.\u003c/p\u003e"},{"header":"METHODS","content":"\u003cp\u003eThis study used the aiHumanoid [\u003cspan class=\"CitationRef\"\u003e12\u003c/span\u003e] v11 simulation system, a high-fidelity fuzzy cognitive map FCM)-based platform designed to model virtual pediatric physiology and disease progression. Seven of the more common pediatric ALL genotypes (mutations) were selected based on known clinical significance and included: TEL-AML1, TEL-AML1\u0026thinsp;+\u0026thinsp;FHx, TEL-AML1\u0026thinsp;+\u0026thinsp;Ikaros Del, TEL-AML1\u0026thinsp;+\u0026thinsp;PAX5 Del, TEL-AML1\u0026thinsp;+\u0026thinsp;Radiation exposure, TEL-AML1\u0026thinsp;+\u0026thinsp;TEL LOH, and TEL-AML1\u0026thinsp;+\u0026thinsp;p16 Del.\u003c/p\u003e\n\u003cp\u003eEach genotype cohort consisted of 25 virtual subjects (n\u0026thinsp;=\u0026thinsp;175 total), aged 1\u0026ndash;20 years (6 cohorts), modeled from the point of achieving remission (T0). Subjects were followed through two post-remission checkpoints: T1 (3 months) and T2 (6 months). All simulation inputs were standardized using the \u0026minus;\u0026thinsp;1 to +\u0026thinsp;1 scale for FCM inputs.\u003c/p\u003e\n\u003cp\u003eTo quantify relapse dynamics within the simulation, two composite metrics were implemented as standalone fuzzy cognitive map (FCM) nodes: the ALL_Relapse Risk Score and the ALL_Relapse Pressure Index (RPI). These metrics respectively estimate snapshot relapse probability and trajectory-based relapse momentum, drawing from weighted input domains related to clonal evolution, minimal residual disease, treatment resistance [\u003cspan class=\"CitationRef\"\u003e7\u003c/span\u003e][\u003cspan class=\"CitationRef\"\u003e10\u003c/span\u003e][\u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e], and immune evasion. A detailed breakdown of input structure, functional behavior, and operational calculation for both metrics is provided in Appendix 2.\u003c/p\u003e\n\u003cp\u003eTable\u0026nbsp;1: The twelve outcome domains were monitored at each timepoint, include:\u003c/p\u003e\n\u003col\u003e\n\u003cli\u003eImmune Activation\u003cbr /\u003e \u0026rarr; The stimulation of immune cells to recognize and respond to threats, including leukemic cells.\u003c/li\u003e\n\u003cli\u003eALL_Clonal Evolution\u003cbr /\u003e \u0026rarr; The stepwise genetic diversification of leukemic cells, leading to more aggressive or therapy-resistant subclones.\u003c/li\u003e\n\u003cli\u003eALL_Development\u003cbr /\u003e \u0026rarr; The biological process through which normal progenitor cells transform into malignant lymphoblasts in acute lymphoblastic leukemia [3][4].\u003c/li\u003e\n\u003cli\u003eALL_Facilitating the Immune Response to ALL\u003cbr /\u003e \u0026rarr; Mechanisms or interventions that enhance immune detection and destruction of ALL cells.\u003c/li\u003e\n\u003cli\u003eALL_Inhibiting the Immune Response to ALL\u003cbr /\u003e \u0026rarr; Processes by which leukemic cells suppress or evade immune surveillance, promoting disease persistence or relapse.\u003c/li\u003e\n\u003cli\u003eALL_Minimal Residual Disease (BC-ALL)\u003cbr /\u003e \u0026rarr; The small number of leukemic cells that remain after treatment in B-cell ALL, undetectable by standard microscopy but detectable by sensitive molecular methods.\u003c/li\u003e\n\u003cli\u003eALL_Pre-leukemic Clone Expansion\u003cbr /\u003e \u0026rarr; The growth of early abnormal cell populations with leukemic potential prior to full malignant transformation.\u003c/li\u003e\n\u003cli\u003eALL_Relapse\u003cbr /\u003e \u0026rarr; The reappearance or progression of leukemia after a period of remission, typically due to treatment-resistant clones.\u003c/li\u003e\n\u003cli\u003eALL_Relapse Pressure Index\u003cbr /\u003e \u0026rarr; A composite measure quantifying the overall biological momentum or stress toward relapse based on longitudinal outcome trends.\u003c/li\u003e\n\u003cli\u003eALL_Relapse Risk Score\u003cbr /\u003e \u0026rarr; A summary metric estimating the likelihood of relapse for a given subject or genotype, derived from key indicators like HLE, MRD, and clonal shifts.\u003c/li\u003e\n\u003cli\u003eALL_Treatment Resistance\u003cbr /\u003e \u0026rarr; The failure of leukemic cells to respond to chemotherapy or immunotherapy, often due to genetic mutations or protective microenvironments.\u003c/li\u003e\n\u003cli\u003eQoL-Physical wellbeing (PW)\u003cbr /\u003e \u0026rarr; A patient\u0026rsquo;s physical health status and ability to perform daily activities, serving as a proxy for treatment burden and recovery.\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eStatistical evaluation used a three-part nonparametric approach:\u003c/p\u003e\n\u003col\u003e\n\u003cli\u003eWilcoxon Signed-Rank Test: Assessed whether the paired differences significantly deviated from zero. The conservative Bonferroni correction was selected to prioritize specificity over sensitivity given the clinical implications of false-positive relapse signals. The corrected p value for statistical significance is P\u0026lt;0.004.\u003c/li\u003e\n\u003cli\u003eCliff\u0026rsquo;s Delta: Quantified the effect size of change, providing population-level directionality and strength of the shift. A Cliff\u0026rsquo;s delta threshold of |\u0026delta;| \u0026ge; 0.474 corresponds to a large effect, chosen to minimize false-positive relapse signals in a high-dimensional longitudinal virtual study.\u003c/li\u003e\n\u003cli\u003eHodges\u0026ndash;Lehmann [21] Estimator (HLE) for paired data: Calculated as the median of all pairwise differences between paired timepoints (T1-T0, T2-T0, and T2-T1) for each outcome and subject, estimating the typical per-subject change.\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003e\u003cbr /\u003e Composite relapse metrics were derived:\u003cbr /\u003e - ALL_Relapse Score:\u003c/p\u003e\n\u003cp\u003e- ALL_Relapse Pressure Index:\u003cbr /\u003e \u003cbr /\u003e All relapse \u0026lsquo;risk\u0026rsquo; metrics in this study represent relative biological propensity within the simulated system, not absolute clinical probability. All results were normalized to the system-standard -1 to +1 scale, and SST (Sober Second Thought) methodology was used for verification.\u003c/p\u003e\n\u003cp\u003eWhile the aiHumanoid simulation platform is proprietary, all model logic relevant to this study\u0026mdash;including outcome domain definitions, composite metric construction, statistical thresholds, and interpretive criteria\u0026mdash;has been fully specified to support independent conceptual replication and methodological scrutiny without requiring access to source code.\u003c/p\u003e"},{"header":"RESULTS","content":"\u003cp\u003eT1 vs T0: Early Recurrence Risk Following Remission in Pediatric Virtual ALL\u003c/p\u003e \u003cp\u003eThis longitudinal in silico study evaluated early recurrence patterns in pediatric acute lymphoblastic leukemia (ALL) [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e] across 175 virtual subjects, each stratified by one of seven TEL-AML1\u0026ndash;based genotypic profiles and six developmental age cohorts (0, 2\u0026ndash;3, 5, 10, 15, and 20 years). Subjects were evaluated across 12 relapse-relevant outcome domains at two time points: T0 (remission) and T1 (3 months post-remission). The Hodges\u0026ndash;Lehmann [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] Estimator (HLE) was used to estimate subject-level shifts, Cliff\u0026rsquo;s delta [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] quantified population-level effect sizes, and the Wilcoxon signed-rank test assessed statistical significance. A Risk Flag was applied when p\u0026thinsp;\u0026lt;\u0026thinsp;0.004 and either HLE\u0026thinsp;\u0026gt;\u0026thinsp;\u0026plusmn;\u0026thinsp;0.1 or |Cliff\u0026rsquo;s delta [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]| \u0026ge; 0.474. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e summarizes the distribution of HLE values across domains and age cohorts for the T1 vs T0 comparison.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eGenotype-Linked Patterns of Early Relapse\u003c/p\u003e \u003cp\u003eAmong the seven genotypes modeled, TEL-AML1\u0026thinsp;+\u0026thinsp;PAX5 Deletion and TEL-AML1\u0026thinsp;+\u0026thinsp;Radiation Exposure emerged as the highest risk groups for early recurrence post-remission.\u003c/p\u003e \u003cp\u003eTEL-AML1\u0026thinsp;+\u0026thinsp;PAX5 Del:\u003c/p\u003e \u003cp\u003eDemonstrated 369 flagged domain events across all age strata.\u003c/p\u003e \u003cp\u003eAt age 10, 9 of 12 domains were flagged, including Clonal Evolution (HLE\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.22, Cliff\u0026rsquo;s d\u0026thinsp;=\u0026thinsp;0.54, p\u0026thinsp;=\u0026thinsp;0.004) and Minimal Residual Disease (MRD) (HLE\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.18, d\u0026thinsp;=\u0026thinsp;0.49, p\u0026thinsp;=\u0026thinsp;0.003).\u003c/p\u003e \u003cp\u003eThis genotype consistently activated early relapse pathways from age 5 onward.\u003c/p\u003e \u003cp\u003eTEL-AML1\u0026thinsp;+\u0026thinsp;Radiation Exposure:\u003c/p\u003e \u003cp\u003eProduced the highest total burden with 379 flagged events.\u003c/p\u003e \u003cp\u003eAt age 15, Treatment Resistance (HLE\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.19, d\u0026thinsp;=\u0026thinsp;0.51, p\u0026thinsp;=\u0026thinsp;0.002) and Clonal Evolution showed strong relapse indicators.\u003c/p\u003e \u003cp\u003eThis genotype exhibited sustained relapse pressure from mid-childhood to early adulthood.\u003c/p\u003e \u003cp\u003eIntermediate-risk genotypes such as TEL-AML1\u0026thinsp;+\u0026thinsp;p16 Deletion and TEL-AML1\u0026thinsp;+\u0026thinsp;Ikaros Deletion showed limited and age-dependent relapse activity. In contrast, TEL-AML1 alone and TEL-AML1\u0026thinsp;+\u0026thinsp;FHx (family history) showed minimal biological deviation from remission, with stable Wilcoxon p-values (p\u0026thinsp;\u0026gt;\u0026thinsp;0.004) and consistently low HLE and Cliff\u0026rsquo;s d values across all ages.\u003c/p\u003e \u003cp\u003eT2 vs T0: Early Relapse Risk at 6 Months Post-Remission in Pediatric ALL\u003c/p\u003e \u003cp\u003eTo investigate the potential for early subclinical relapse, we analyzed virtual longitudinal profiles comparing T2 (6 months post-remission) with T0 (remission baseline) across pediatric acute lymphoblastic leukemia [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e][\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e] (ALL) subjects. This comparison was conducted using three complementary statistical approaches: the Hodges\u0026ndash;Lehmann [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] Estimator (HLE) for subject-level shifts, Cliff\u0026rsquo;s delta [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] for population-level effect sizes, and the Wilcoxon signed-rank test for statistical significance. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents the HLE estimates for T2 vs T0 across the 12 relapse domains and six age cohorts.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWe defined high-risk relapse strata as those that met the following criteria: Wilcoxon p\u0026thinsp;\u0026lt;\u0026thinsp;0.004, |HLE| \u0026gt; 0.1, and |Cliff\u0026rsquo;s delta [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]| \u0026ge; 0.474. This composite model identifies both clinically relevant and statistically robust relapse signals.\u003c/p\u003e \u003cp\u003eOut of 72 possible age \u0026times; outcome combinations (6 age cohorts \u0026times; 12 outcome domains), 18 were flagged as high risk (25%). The most pronounced risk signals were seen in the 2-3-year and 5-year age cohorts, which together accounted for 11 of the 18 flagged combinations.\u003c/p\u003e \u003cp\u003eAge-Dependent Patterns: The 2-3-year and 5-year cohorts showed the highest burden of relapse risk, with major contributions from clonal evolution, minimal residual disease (MRD), relapse pressure index, and treatment resistance [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e][\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e][\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. The 10-year group also showed moderate relapse activity, while ages 0 and 20 were relatively stable.\u003c/p\u003e \u003cp\u003eDomain-Specific Insights: Among the 12 relapse-related domains, four were especially prominent in driving high-risk profiles:\u003c/p\u003e \u003cp\u003eALL_Clonal Evolution: Frequently flagged across ages 2\u0026ndash;3 to 10 with HLE\u0026thinsp;\u0026gt;\u0026thinsp;0.14 and Cliff\u0026rsquo;s delta [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u0026thinsp;\u0026gt;\u0026thinsp;0.50.\u003c/p\u003e \u003cp\u003eALL_Minimal Residual Disease (BC-ALL): Persistent positivity in age 5 cohort (HLE\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.230, Cliff\u0026rsquo;s d\u0026thinsp;=\u0026thinsp;0.51, Wilcoxon p\u0026thinsp;=\u0026thinsp;0.002).\u003c/p\u003e \u003cp\u003eALL_Treatment Resistance: Flagged in ages 5\u0026ndash;15, suggesting reemergence of drug-resistant subclones.\u003c/p\u003e \u003cp\u003eALL_Relapse Pressure Index: Increased systemic stress in leukemic niches was detected in ages 2\u0026ndash;3\u0026ndash;10.\u003c/p\u003e \u003cp\u003eIllustrative High-Risk Stratum: At age 5, Minimal Residual Disease (MRD) showed a subject-level increase (HLE\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.230), a large effect size (Cliff\u0026rsquo;s d\u0026thinsp;=\u0026thinsp;0.51), and strong statistical support (Wilcoxon p\u0026thinsp;=\u0026thinsp;0.002), fulfilling all three criteria for high-risk flagging.\u003c/p\u003e \u003cp\u003eT2 vs T1: Late Subclinical Recurrence Risk at 6 Months vs 3 Months Post-Remission\u003c/p\u003e \u003cp\u003eTo assess continued relapse risk trajectories after initial post-remission surveillance, we compared virtual pediatric ALL profiles between T2 (6 months post-remission) and T1 (3 months post-remission) across 175 subjects. Using Hodges\u0026ndash;Lehmann [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] Estimator (HLE), Cliff\u0026rsquo;s delta [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], and Wilcoxon signed-rank test, we identified relapse-relevant changes at the subject level, population level, and statistical level, respectively. We defined high-risk strata as those with Wilcoxon p\u0026thinsp;\u0026lt;\u0026thinsp;0.004, |HLE| \u0026gt; 0.1, and |Cliff\u0026rsquo;s delta [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]| \u0026ge; 0.474.\u003c/p\u003e \u003cp\u003eOut of 72 age \u0026times; domain strata, 21 were flagged as high risk (29.2%), indicating an increased burden of biological recurrence activity in this interval compared to earlier comparisons. The 2-3-year, 5-year, and 10-year age cohorts were especially vulnerable, contributing 16 of the 21 high-risk strata (76%).\u003c/p\u003e \u003cp\u003eAge-Stratified Patterns\u003c/p\u003e \u003cp\u003e2\u0026ndash;3 Years: Clonal Evolution, MRD, and Treatment Resistance continued to show increasing divergence (e.g., MRD HLE\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.19, Cliff\u0026rsquo;s d\u0026thinsp;=\u0026thinsp;0.50, p\u0026thinsp;=\u0026thinsp;0.002).\u003c/p\u003e \u003cp\u003e5 Years: Relapse Pressure Index and Resistance metrics showed strong relapse signaling.\u003c/p\u003e \u003cp\u003e10 Years: A resurgence of Clonal Evolution and Resistance domains reflected cumulative relapse burden.\u003c/p\u003e \u003cp\u003eOlder ages (15, 20) exhibited partial stabilization, with only 2 domains flagged in total, suggesting possible biological resolution or treatment buffering.\u003c/p\u003e \u003cp\u003eDomain-Specific Insights\u003c/p\u003e \u003cp\u003eThree domains remained most predictive of recurrence risk in this later window:\u003c/p\u003e \u003cp\u003eALL_Clonal Evolution: Persistent across 2\u0026ndash;3 to 10 years, with several strata exceeding Cliff\u0026rsquo;s d\u0026thinsp;=\u0026thinsp;0.50 and HLE\u0026thinsp;\u0026gt;\u0026thinsp;+\u0026thinsp;0.15.\u003c/p\u003e \u003cp\u003eALL_Minimal Residual Disease (BC-ALL): Especially in 2\u0026ndash;3 and 5 years, indicating persistent leukemic cell populations.\u003c/p\u003e \u003cp\u003eALL_Treatment Resistance: Suggests evolving therapy escape mechanisms in intermediate ages.\u003c/p\u003e \u003cp\u003eOne notable addition was subtle emergence of ALL_Relapse Risk Score in older children (age 10\u0026ndash;15), with HLE\u0026thinsp;\u0026gt;\u0026thinsp;+\u0026thinsp;0.11 and Cliff\u0026rsquo;s d\u0026thinsp;\u0026gt;\u0026thinsp;0.45, though not reaching all criteria thresholds (p\u0026thinsp;\u0026gt;\u0026thinsp;0.004).\u003c/p\u003e \u003cp\u003eHigh-Risk Stratum Example\u003c/p\u003e \u003cp\u003eAt age 2\u0026ndash;3, MRD was again prominently elevated (HLE\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.19, Cliff\u0026rsquo;s d\u0026thinsp;=\u0026thinsp;0.50, Wilcoxon p\u0026thinsp;=\u0026thinsp;0.002), meeting all flagging thresholds and reinforcing its role as a core recurrence driver.\u003c/p\u003e \u003cp\u003eSummary: Risk Stratification of Early and Subclinical Recurrence in Pediatric ALL\u003c/p\u003e \u003cp\u003eAnalysis across three post-remission intervals (T0 vs T1, T0 vs T2, and T2 vs T1) identified consistent genotype- and age-specific patterns of relapse vulnerability [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e][\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e] in pediatric acute lymphoblastic leukemia [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e][\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e] (ALL). Subjects were classified into high, intermediate, or low risk categories based on the number of flagged relapse domains, strength of statistical signals (p\u0026thinsp;\u0026lt;\u0026thinsp;0.004, HLE\u0026thinsp;\u0026gt;\u0026thinsp;0.1, |Cliff\u0026rsquo;s d| \u0026ge; 0.474), and persistence across time points.\u003c/p\u003e \u003cp\u003eHigh-Risk Genotypes\u003c/p\u003e \u003cp\u003eThese subtypes showed recurrent, multi-domain relapse activation across all three-time intervals, particularly during mid-childhood:\u003c/p\u003e \u003cp\u003eTEL-AML1\u0026thinsp;+\u0026thinsp;PAX5 Deletion\u003c/p\u003e \u003cp\u003ePersistent MRD elevation, Clonal Evolution, and Treatment Resistance from age 5 onward.\u003c/p\u003e \u003cp\u003eUp to 9 of 12 relapse domains flagged at age 10.\u003c/p\u003e \u003cp\u003eConsistent early and late recurrence indicators.\u003c/p\u003e \u003cp\u003eTEL-AML1\u0026thinsp;+\u0026thinsp;Radiation Exposure\u003c/p\u003e \u003cp\u003eHighest total number of flagged domains.\u003c/p\u003e \u003cp\u003eSustained relapse signaling in Treatment Resistance and Clonal Evolution from ages 10 to 15.\u003c/p\u003e \u003cp\u003eStrong relapse pressure despite initial remission.\u003c/p\u003e \u003cp\u003eRecommended action: Intensified post-remission surveillance, especially between ages 5\u0026ndash;15.\u003c/p\u003e \u003cp\u003eIntermediate-Risk Genotypes\u003c/p\u003e \u003cp\u003eThese groups showed limited or age-dependent recurrence features, often confined to one time point or fewer domains:\u003c/p\u003e \u003cp\u003eTEL-AML1\u0026thinsp;+\u0026thinsp;p16 Deletion\u003c/p\u003e \u003cp\u003eTEL-AML1\u0026thinsp;+\u0026thinsp;Ikaros Deletion\u003c/p\u003e \u003cp\u003eDemonstrated relapse signatures in selected age strata (typically 5\u0026ndash;10 years).\u003c/p\u003e \u003cp\u003ePartial but non-persistent elevation of MRD or Clonal Evolution.\u003c/p\u003e \u003cp\u003eRecommended action: Age-specific follow-up, especially during mid-childhood.\u003c/p\u003e \u003cp\u003eLow-Risk Genotypes\u003c/p\u003e \u003cp\u003eThese profiles remained biologically stable across all time points with negligible domain flagging:\u003c/p\u003e \u003cp\u003eTEL-AML1 Alone (Baseline)\u003c/p\u003e \u003cp\u003eTEL-AML1\u0026thinsp;+\u0026thinsp;FHx (Family History)\u003c/p\u003e \u003cp\u003eNo significant domain activation (Wilcoxon p\u0026thinsp;\u0026gt;\u0026thinsp;0.004 across all comparisons).\u003c/p\u003e \u003cp\u003eConsistently low HLE and Cliff\u0026rsquo;s delta [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] metrics.\u003c/p\u003e \u003cp\u003eRecommended action: Standard surveillance schedules appear sufficient.\u003c/p\u003e \u003cp\u003eAge-Based Recurrence Risk Patterns\u003c/p\u003e \u003cp\u003eRelapse vulnerability exhibited a developmentally stratified pattern that interacted with genotype-specific risk:\u003c/p\u003e \u003cp\u003e\u0026bull; Highest Risk Ages:\u003c/p\u003e \u003cp\u003eAges 2\u0026ndash;3, 5, and 10 years demonstrated the greatest burden of flagged relapse domains\u0026mdash;particularly MRD, Clonal Evolution, and Treatment Resistance\u0026mdash;in high-risk genotypes (e.g., PAX5 Del, Radiation).\u003c/p\u003e \u003cp\u003e\u0026bull; Lowest Risk Ages:\u003c/p\u003e \u003cp\u003eInfants (age 0) and older adolescents (age 20) consistently showed minimal relapse signal across genotypes, suggesting relative resilience or transient therapeutic sensitivity.\u003c/p\u003e \u003cp\u003eThis developmental gradient reinforces the need for age-stratified surveillance, particularly during mid-childhood in genetically high-risk patients.\u003c/p\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eThis study expands upon our previously published virtual modeling framework for childhood ALL by introducing a longitudinal, genotype- and age-stratified analysis of post-remission relapse risk. Using aiHumanoid [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] v11.0, we simulated disease trajectories across 175 virtual pediatric subjects, incorporating seven clinically relevant TEL-AML1-based genotypes and six developmental age cohorts. Outcomes were evaluated across 12 biologically grounded relapse domains at three key time points (T0, T1, T2).\u003c/p\u003e \u003cp\u003eThe central innovation of this work lies in the combined use of the Hodges\u0026ndash;Lehmann [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] Estimator (HLE), Cliff\u0026rsquo;s delta [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], and the Wilcoxon signed-rank test. Together, these metrics enabled high-resolution detection of both subject-level and population-level shifts in relapse risk. HLE proved valuable in quantifying early subclinical changes that may precede overt clinical relapse.\u003c/p\u003e \u003cp\u003eRisk Stratification by Genotype and Age\u003c/p\u003e \u003cp\u003eA key insight from this analysis is the emergence of distinct risk tiers based on genotype:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eHigh-risk genotypes (TEL-AML1\u0026thinsp;+\u0026thinsp;PAX5 Deletion and TEL-AML1\u0026thinsp;+\u0026thinsp;Radiation Exposure) show consistent, multi-domain relapse activation across all time points, with up to 9 of 12 outcome domains flagged in mid-childhood cohorts. These genotypes represent urgent candidates for intensified post-remission surveillance and potentially early intervention trials.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eIntermediate-risk genotypes (TEL-AML1\u0026thinsp;+\u0026thinsp;Ikaros Deletion and TEL-AML1\u0026thinsp;+\u0026thinsp;p16 Deletion) demonstrated transient or age-restricted relapse signals, particularly between ages 5\u0026ndash;10. While less aggressive than high-risk profiles, they may still benefit from tailored monitoring strategies during key developmental windows.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eLow-risk genotypes (TEL-AML1 alone and TEL-AML1\u0026thinsp;+\u0026thinsp;FHx) remained biologically stable throughout, with negligible HLE shifts and consistently non-significant Wilcoxon p-values. These subgroups may be candidates for reduced-intensity follow-up, limiting unnecessary exposure to monitoring-associated burdens.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eDevelopmental Windows of Vulnerability\u003c/p\u003e \u003cp\u003eImportantly, relapse risk also demonstrated a developmentally stratified pattern, with the ages 2\u0026ndash;3, 5, and 10 years consistently associated with the highest burden of flagged relapse domains, independent of genotype. These ages likely reflect critical intersections of immune reconstitution, bone marrow microenvironmental dynamics, and therapy responsiveness. In contrast, infants (age 0) and older adolescents (age 20) consistently showed biological stability, suggesting temporary windows of resilience.\u003c/p\u003e \u003cp\u003eThis dual-axis stratification (by genotype and age) provides a powerful framework for anticipating relapse dynamics and refining post-remission care. The visual trends shown in Figs.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e reinforce the emergence of MRD, Clonal Evolution, and Treatment Resistance as dominant relapse drivers [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e][\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e][\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e] during mid-childhood.\u003c/p\u003e \u003cp\u003eIn contrast to most \u0026ldquo;digital twin\u0026rdquo; implementations that rely primarily on static predictive models or correlative machine learning [14,15,25], the aiHumanoid framework employed here is explicitly causal, physiologically grounded, and dynamically adaptive. The system models disease progression as an evolving network of interacting biological processes, enabling inference about relapse momentum, temporal escalation, and counterfactual trajectories rather than point estimates alone. This distinction is particularly important in post-remission surveillance, where early biological drift may precede measurable clinical relapse. By capturing trajectory-level dynamics across genotype and developmental stage, this approach extends beyond risk scoring toward anticipatory disease modeling.\u003c/p\u003e \u003cp\u003eSimulation Metrics and Clinical Relevance\u003c/p\u003e \u003cp\u003eThe derived composite indices\u0026mdash;\u003cb\u003eALL_Relapse Score\u003c/b\u003e and \u003cb\u003eRelapse Pressure Index (RPI)\u003c/b\u003e\u0026mdash;proved effective in summarizing relapse dynamics. When paired with primary markers such as MRD and Clonal Evolution, these metrics enable a structured approach to early warning detection. With further validation, they could form the basis of real-time clinical dashboards powered by patient-specific genomic and treatment data.\u003c/p\u003e \u003cp\u003eLimitations and Future Directions\u003c/p\u003e \u003cp\u003eThis work is grounded in high-fidelity, physiology-informed virtual simulations. However, these findings should be interpreted with caution until validated against real-world longitudinal datasets. Future work will focus on calibrating simulation-derived risk metrics with clinical relapse outcomes and extending the model to include adaptive therapies or immune-targeted strategies.\u003c/p\u003e \u003cp\u003eAs longitudinal pediatric relapse datasets mature and federated data-sharing initiatives expand, prospective calibration of simulation-derived relapse metrics against real-world outcomes will become increasingly feasible.\u003c/p\u003e \u003cp\u003eReproducibility Statement\u003c/p\u003e \u003cp\u003eWhile the aiHumanoid simulation engine is proprietary, all causal structures, outcome domains, composite metric definitions, thresholds, and statistical criteria required for independent conceptual replication are fully specified in this manuscript. Independent validation can be performed using alternative causal simulation frameworks implementing equivalent domain logic.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis virtual longitudinal study demonstrates the potential feasibility and utility of combining HLE, Cliff\u0026rsquo;s delta [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], and Wilcoxon testing to quantify genotype-specific relapse risk in childhood ALL. The results identify high-risk and low-risk genotypes early in the surveillance period, offering a pathway for precision follow-up care.\u003c/p\u003e \u003cp\u003eOur approach supports individualized post-remission planning, provides deeper insights into the biological underpinnings of relapse, and establishes a scalable method for evaluating new genomic variants. The integration of advanced statistical tools and simulation modeling offers a powerful paradigm for anticipating and managing relapse in pediatric oncology.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eCompeting interests\u003c/p\u003e\n\u003cp\u003eWayne R. Danter is the Founder and developer of the aiHumanoid technology and serves as Chief Scientific Officer of HumanQAI, Inc. (formerly 123Genetix, Inc.). All other authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003eFunding\u003c/p\u003e\n\u003cp\u003eThis research received no external funding.\u003c/p\u003e\n\u003cp\u003eEthics approval and consent to participate\u003c/p\u003e\n\u003cp\u003eNot applicable. This study was conducted entirely in silico using population-based virtual pediatric acute lymphoblastic leukemia models. No human participants, patient-level clinical data, or identifiable biological specimens were involved; therefore, ethics approval and informed consent were not required.\u003c/p\u003e\n\u003cp\u003eConsent for publication\u003c/p\u003e\n\u003cp\u003eNot applicable. No human participants or identifiable data were included in this study.\u003c/p\u003e\n\u003cp\u003eData availability\u003c/p\u003e\n\u003cp\u003eThe datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request. Data are not publicly available due to the proprietary nature of the aiHumanoid simulation platform.\u003c/p\u003e\n\u003cp\u003eAuthor contributions\u003c/p\u003e\n\u003cp\u003eConceptualization: W.R.D.; Methodology: W.R.D.; Formal analysis: W.R.D.; Writing—original draft: W.R.D.; Writing—review and editing: all authors.\u003c/p\u003e\n\u003cp\u003eAcknowledgements\u003c/p\u003e\n\u003cp\u003eThe authors acknowledge limited use of GPT-5.1 (OpenAI) as a supporting tool in the development and refinement of population-based virtual subject representations within the aiHumanoid simulation framework. GPT-5.1 was not used to generate real patient data, to perform statistical analyses, or to make independent scientific or clinical decisions. All model design choices, simulation parameters, data interpretation, and conclusions were defined, reviewed, and approved by the authors, who retain full responsibility for the work.\u003c/p\u003e\n\u003cp\u003eClinical trial registration\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eFarber S, Diamond LK, Mercer RD, Sylvester RF, Wolff JA. Temporary remissions in acute leukemia in children produced by folic acid antagonist, 4-aminopteroyl-glutamic acid (aminopterin). \u003cem\u003eN Engl J Med.\u003c/em\u003e 1948;238(23):787\u0026ndash;93.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePinkel D, Simone JV, Hustu HO, Aur RJ. Total therapy of acute lymphocytic leukemia. \u003cem\u003ePediatrics.\u003c/em\u003e 1971;48(5):846\u0026ndash;57.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eInaba H, Greaves M, Mullighan CG. Acute lymphoblastic leukaemia. \u003cem\u003eLancet\u003c/em\u003e. 2013;381(9881):1943\u0026ndash;55.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePui CH, Yang JJ, Hunger SP, Pieters R, Schrappe M, Biondi A, et al. Childhood acute lymphoblastic leukemia: progress through collaboration. \u003cem\u003eJ Clin Oncol.\u003c/em\u003e 2015;33(27):2938\u0026ndash;48.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOskarsson T, S\u0026ouml;derh\u0026auml;ll S, Arvidson J, Forestier E, Montgomery S, Frandsen TL, et al. Relapsed childhood acute lymphoblastic leukaemia in the Nordic countries: risk factors, treatment and outcome. \u003cem\u003eHaematologica.\u003c/em\u003e 2016;101(1):68\u0026ndash;76.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCeppi F, Cazzaniga G, Colombini A, Biondi A, Conter V, et al. Risk factors for relapse in childhood acute lymphoblastic leukemia. \u003cem\u003eLeuk Lymphoma.\u003c/em\u003e 2015;56(6):1907\u0026ndash;18.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHughes A, Wade R, Richards S, Hancock J, Vora A, Mitchell C, et al. Treatment and outcome of relapsed childhood T-cell acute lymphoblastic leukaemia: results from the UKALLR3 trial. \u003cem\u003eHaematologica.\u003c/em\u003e 2020;105(3):574\u0026ndash;82.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRheingold SR, Salzer WL, Devidas M, Borowitz MJ, Carroll AJ, Gastier-Foster JM, et al. Determinants of survival after first relapse of acute lymphoblastic leukemia. \u003cem\u003eLeukemia.\u003c/em\u003e 2024;38(2):456\u0026ndash;68.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDel Rocco NJ, Annesley CE, Alexander TB, Bartram J, Attarbaschi A, Moorman AV, et al. Enhanced risk stratification for children and young adults with ALL: the PI-UKALL score. \u003cem\u003eLeukemia.\u003c/em\u003e 2024;38(6):1231\u0026ndash;42.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEckert C, von Stackelberg A, Seeger K, Groeneveld-Krentz S, Cario G, Schrauder A, et al. Risk factors and outcomes in children with high-risk B-cell acute lymphoblastic leukemia. \u003cem\u003eCancer Lett.\u003c/em\u003e 2021;499:154\u0026ndash;64.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSlayton WB, Devidas M, Hunger SP, Heerema NA, Carroll AJ, Winick NJ, et al. Poverty and relapse risk in children with acute lymphoblastic leukaemia. \u003cem\u003eBlood.\u003c/em\u003e 2023;142(3):221\u0026ndash;32.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDanter WR. Investigating genetic and familial risks in childhood ALL: a longitudinal virtual study using aiHumanoid simulations of TEL-AML1 gene fusion. \u003cem\u003emedRxiv\u003c/em\u003e. 2024. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1101/2024.06.19.24309162\u003c/span\u003e\u003cspan address=\"10.1101/2024.06.19.24309162\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKatsoulakis E, Cloutier M, Vasan A, Bonnet A, Lemaire J, Shah NR, et al. Digital twins for health: a scoping review. \u003cem\u003eNPJ Digit Med.\u003c/em\u003e 2024;7:110.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSad\u0026eacute;e C, Brochard J, van der Garde M, Chouvarda I, Garcia-Remesal M, Roca J, et al. Medical digital twins: enabling precision medicine and personalized healthcare delivery. \u003cem\u003eLancet Digit Health.\u003c/em\u003e 2025;7(2):e123-34.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGiansanti D, Simeoni R, Barcellona A, Mercuri NB, Morabito FC, et al. Exploring the potential of digital twins in cancer treatment. \u003cem\u003eJ Clin Med.\u003c/em\u003e 2025;14(10):3574.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhuang Y, Zhao J, Li H, Zhang T, Sun H, et al. Advancing precision oncology with digital and virtual twins. \u003cem\u003eFront Oncol.\u003c/em\u003e 2025;15:119\u0026ndash;32.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi X, Zhou Y, Chen Q, Wang J, Liu S, et al. From virtual to reality: innovative practices of digital twins in tumor therapy. \u003cem\u003eJ Transl Med.\u003c/em\u003e 2025;23:117.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDemir A, Chen Y, G\u0026oacute;mez J, Zhou H, Lee J, et al. Towards verifiable cancer digital twins: tissue-level modelling and beyond. \u003cem\u003eFront Physiol.\u003c/em\u003e 2024;15:1473125.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBruynseels K, Santoni de Sio F, van den Hoven J. Exploring approaches for predictive cancer patient digital twins. \u003cem\u003eFront Oncol.\u003c/em\u003e 2020;10:9586248.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCliff N. Dominance statistics: ordinal analyses to answer ordinal questions. \u003cem\u003ePsychol Bull.\u003c/em\u003e 1993;114(3):494\u0026ndash;509.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHodges JL, Lehmann EL. Estimates of location based on rank tests. \u003cem\u003eAnn Math Stat.\u003c/em\u003e 1963;34(2):598\u0026ndash;611.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim, H., Yoon, S.H., Kang, S. \u003cem\u003eet al.\u003c/em\u003e Improved survival in pediatric acute lymphoblastic leukemia through therapy intensification based on minimal residual disease and protocol-driven early response risk classification. \u003cem\u003eBlood Res.\u003c/em\u003e 60, 40 (2025). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s44313-025-00085-3\u003c/span\u003e\u003cspan address=\"10.1007/s44313-025-00085-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eȘtefan A-I, Radu L-E, Jardan D and Coliță A (2025) The emerging role of next-generation sequencing in minimal residual disease assessment in acute lymphoblastic leukemia: a systematic review of current literature. \u003cem\u003eFront. Med\u003c/em\u003e. 12:1570041. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fmed.2025.1570041\u003c/span\u003e\u003cspan address=\"10.3389/fmed.2025.1570041\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMenon H, Singh PK, Bagal B, Dolai T, Jain A, Chaudhri A. Minimal Residual Disease in the Management of B-Cell Acute Lymphoblastic Leukemia: A Systematic Review of Studies from Indian Settings. Indian J Hematol Blood Transfus. 2024;40(1):1\u0026ndash;11. doi: 10.1007/s12288-023-01641-6. Vall\u0026eacute;e A. Digital Twins for Personalized Medicine Require Epidemiological Data and Mathematical Modeling: Viewpoint J Med Internet Res 2025;27:e72411 doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2196/72411\u003c/span\u003e\u003cspan address=\"10.2196/72411\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNCI PDQ (2025)-PDQ\u0026reg; Pediatric Treatment Editorial Board. PDQ Childhood Acute Lymphoblastic Leukemia Treatment. Bethesda, MD: National Cancer Institute. 2025, Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.cancer.gov/types/leukemia/hp/child-all-treatment-pdq\u003c/span\u003e\u003cspan address=\"https://www.cancer.gov/types/leukemia/hp/child-all-treatment-pdq\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Acute Lymphoblastic Leukemia, Pediatric Oncology, aiHumanoid, Hodges–Lehmann Estimator, Genetic Relapse Risk, Simulation Study, TEL-AML1, PAX5, Radiation","lastPublishedDoi":"10.21203/rs.3.rs-8653812/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8653812/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cb\u003eBackground\u003c/b\u003e\u003c/p\u003e \u003cp\u003eDespite advances in front-line therapy for pediatric acute lymphoblastic leukemia (ALL), relapse remains a leading cause of treatment failure and mortality. Building on prior virtual investigations of TEL-AML1\u0026ndash;associated leukemogenesis, this study evaluates post-remission relapse risk across pediatric ALL genotypes using mechanistic simulation.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMethods\u003c/b\u003e\u003c/p\u003e \u003cp\u003eSeven clinically relevant pediatric ALL genotypes were modeled using aiHumanoid v11.0 simulations. Each cohort was simulated from remission (T0) through two post-remission intervals (T1\u0026thinsp;=\u0026thinsp;3 months; T2\u0026thinsp;=\u0026thinsp;6 months). Longitudinal change was assessed using the Hodges\u0026ndash;Lehmann Estimator (HLE), Wilcoxon signed-rank test, and Cliff\u0026rsquo;s delta. Composite relapse metrics were derived from 12 relapse-related outcome domains.\u003c/p\u003e\u003cp\u003e\u003cb\u003eResults\u003c/b\u003e\u003c/p\u003e \u003cp\u003eEarly relapse signals were detected in high-risk genotypes, including TEL-AML1\u0026thinsp;+\u0026thinsp;PAX5 deletion, TEL-AML1\u0026thinsp;+\u0026thinsp;radiation exposure, and TEL-AML1\u0026thinsp;+\u0026thinsp;Ikaros deletion, characterized by elevated HLE values and large effect sizes. Low-risk genotypes, including TEL-AML1 alone and TEL-AML1\u0026thinsp;+\u0026thinsp;family history, showed minimal deviation from remission through 6 months.\u003c/p\u003e\u003cp\u003e\u003cb\u003eConclusions\u003c/b\u003e\u003c/p\u003e \u003cp\u003eHLE-based longitudinal analysis enables detection of genotype-specific relapse dynamics in pediatric ALL and supports individualized, time-aware post-remission surveillance strategies.\u003c/p\u003e","manuscriptTitle":"Genotype- and Age-Specific Relapse Dynamics in Childhood Acute Lymphoblastic Leukemia: A Longitudinal Mechanistic Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-25 10:23:47","doi":"10.21203/rs.3.rs-8653812/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"8035c8d1-e0a4-44ed-839f-fbdf46f384b0","owner":[],"postedDate":"January 25th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":61522942,"name":"Health sciences/Medical research/Preclinical research"},{"id":61522943,"name":"Biological sciences/Biotechnology/Genomics/Pharmacogenomics"}],"tags":[],"updatedAt":"2026-02-05T14:57:22+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-25 10:23:47","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8653812","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8653812","identity":"rs-8653812","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}