STATISTICAL ANALYSIS OF BIOPHYSICAL PROPERTIES IN CELLS FOR LEUKEMIA PREDICTIVE MODELS

Abstract

Improving life expectancy presents challenges, including eradicating fatal diseases like childhood and adolescent leukemia. Such pathophysiological aggressiveness arises into malignant phenotypes, leading to treatment-resistant and metastatic situations due to mechanical adaptive and migratory characteristics. This work investigates on the mechanical properties of various leukemic tumor cells compared to healthy non-tumoral cells, quantified as the Mean Square Displacement (MSD) for each cell chromatin grain. The goal is to identify biophysical markers for early diagnosis and treatment.

To achieve our objective, we based the operational research on two statistical methods: (1) the non-parametric hypothesis test for mean differences, applied by means of Bootstrap, and (2) the Kullback-Leibler divergence by estimating the distribution of biophysical markers' statistics by Parzen windows. Our analysis is based on the hypothesis that the biophysical properties of cells, as determined by their functional phenotype, influence the invasive and aggressive development of cancer cells. Hence, identifying and understanding these features is critical for the diagnosis and effective treatment of cancer diseases. We compare nuclear mechanical properties in tumour and healthy cells through the analysis of seven features, derived from the chromatin grain MSD: mean, median, standard deviation, heterogeneity, skewness, kurtosis and diffusivity. For this purpose, a multiple particle tracking analysis for the chromatin grains is performed.

Our analysis identify promising differential biomarkers between different tumour and healthy cells, highlighting the potential of this approach in the early detection or relapse and metastasis in leukaemia patients. Most relevant features for discriminant analysis are the mean, the median and the standard deviation. Lines for improvement are identified and future research directions are proposed.