Model-aided analysis identifies common HSPC dysregulation in myeloid neoplasms. (A) CD34⁺ HSPC scRNA-seq data from patients with MDS and healthy controls (HC) (B: patients with MPN and HC) projected into the same gene expression space as the murine models. (C) Expression of NFκB-response genes, Nfkbia and Irf1, in MDS vs HC HSPCs (D: MPN vs HC HSPCs). Italicized number is P value given by Wilcoxon rank-sum test. (E) Ratio of MDS to HC HSPC density (F: MPN to HC HSPC density) over pseudotime (assuming equal number of HSPCs in each condition) estimated 1000 times to reflect uncertainty in lineage assignments; solid line indicates median and shaded region covers minimum to maximum value. (G) Difference in branching ratio between MDS (solid lines) or MPN (dashed lines) and HC inferred from fitting the population dynamics model to the scRNA-seq data with different branch-point values. Lines connect median values and error bars denote maximum/minimum values across 500 fits. (H) Analogous to G for the average difference of developmental-flux functions in early (branch-point to 0.5) progenitor lineages. (I) Expression of Cebpa and Ikzf1 along the HSC lineage in MDS patients (left) and MPN patients (right) compared with healthy controls. (J) Expression of Myc and c-Myc response gene, Cdk4, along the MyP lineage in patients with MDS (left) and patients with MPN (right) compared with healthy controls. Lines in panels I-J are NB-GAM fits, shaded regions are 95% confidence intervals, and transparent points are individual cell data. Italicized numbers are P values and degrees of freedom used for NB-GAM fits specified by k.