Fig. 2: Patient characteristics that affect the model goodness of fit. | npj Digital Medicine

Fig. 2: Patient characteristics that affect the model goodness of fit.

From: Digital twin models for predicting venetoclax and azacitidine-induced neutropenia in patients with acute myeloid leukemia

Fig. 2

a Assumption 1 (A1)—this shows the correlation between time elapsed since the end of the last treatment cycle and the neutrophil counts, among the 71 patients used for modeling. b Assumption 2 (A2)—correlation between interpolated bone marrow blast percentage and neutrophil counts for the same 71 patients. c Fraction of variance unexplained for all models, separated into patients for whom A1 is true and patients for whom A1 is false. d Same as previous, but for Assumption 2. e The actual FVU vs. predicted FVU for M2c, where the predicted FVU was calculated using a LASSO linear regression model based on the clinical and genomic features upon patient intake. f The top positive and negative coefficients for the previous LASSO model. Positive coefficients indicate that a variable is predictive of a high FVU, while negative coefficients are the opposite.

Back to article page