Previously, a deductive method was used to model deviation from Mendelian polygenic inheritance to reveal a model with up to 5 levels of gene expression in polygenic combinations. Multilevel polygenic expression leads to phenotypes with specific prevalence levels. Four levels of prevalence were reliably simulated with the lowest prevalence being due to dominant impact from single genes. Following this deductive conclusion, an Exploratory pilot study demonstrated probable capacity to predict impact on gene expression from 3D conformation and histone modification from genome browser data. A plausible model is proposed of interaction between inflammation and expression of genes relevant for mental health in cancer, which simultaneously explains the multilevel gene expression hypothesis. Next, a recently developed AI algorithm LEON (Life Emulating Orchestrated Net) was run on a workstation to recognise topological patterns in genome data. Preliminary application of this AI method is able to accurately identify Histone modification marks associated with Euchromatin status with an accuracy of 95-98%. Such topological analysis with resolution at the level of enhancer-promoter loops (30-50 Kb) found differences between control (GM12878) and pathological (Leukemia in blast crisis – K562) lymphoblastic cell lines. Harnessing the ability to match topological data with histone marks could generate models of gene-environment interactions using the multi-level expression model. Given the similarity of inflammatory status t is proposed to use this form of modelling to predict intervention strategies to mitigate ICANS in CART-T.