《学术论文集锦[开错]》紫微客

1、001

　　Predict responsiveness to chemotherapy for Hepatoblastoma by statistical inference on the vitro trials of patient-derived tumor organoids

　　1.Introduction

　　Hepatoblastoma (HB) is the most common malignant tumor of liver in infants and young children. It is a serious harm to the health of children. 90% of it occurs before the age of 3, and 60% of it is infants under the age of 1[1]. Therefore, even though HB is a rare disease in the whole population, it still attracts high attention from the medical community. Multi-drug chemotherapy and surgery are the main means to treat HB, but individualized precision medicine cannot be achieved at present. Therefore, in order to effectively screen and apply antitumor drugs, it is crucial to guide individual chemotherapy through tumor chemosensitive assay in vitro[1].

　　Tumor chemosensitive assay in vitro is an important preclinical study, which aims to guide clinical drug use by measuring the relationship between drug concentration and cell response[2]. The basic procedure is to first expose cells from the patient to a specific dose of the drug, and then measure the number of living cells (or fluorescence intensity, SI) at a certain point in time (for example, five days later)[3]. At present, the mainstream dose-response research uses the data obtained from the experiment to carry out curve-fitting, which is used to calculate the relevant evaluation indicators of dose-response, such as the maximal effect (Emax), the half maximal effective concentration (EC50) and the half maximal inhibitory concentration (IC50), etc.[2]. However, differences in cell division frequency will affect drug resistance. Therefore, in order to make the comparison of drug efficacy reliable, scholars used normalized growth rate (GR values) inhibition to replace the above indicators, because studies had shown that GR values are basically not affected by cell division rate[4]. For example, the concentration at which GR(c) = 0.5 (GR50) and the half of its maximal effect on cell growth (GEC50) obtained from GR dose-response curves are similar to IC50 and EC50, etc.[3].

　　However, for rare diseases such as HB, the high cost and small number of patients involved in the tumor chemosensitive assay in vitro means that many classical statistical analysis methods are not applicable. Therefore, in order to provide reliable drug guidance for clinical trials, we must use statistical inference methods of small samples to conduct statistical analysis of the data obtained from tumor chemosensitive assay in vitro[5]. At present, the method of small sample statistical analysis has been applied in many fields. For example, the Bayes theory of small sample statistics was applied to evaluate the maintenance effect of large navigation equipment[6]; Bayes algorithm was used to establish a dynamic recursive model to estimate the qualified rate of mechanical products[7]; For a class of high-dimensional small sample data, variable selection based on Wilcoxon rank sum test and variable compression based on Partial least squares (PLS) were implemented to reduce dimension[8]; Using virtual sample generation technology to improve the learning performance of small sample datasets[9]; The Bootstrap method was used to improve the DEA method for the analysis of energy efficiency in BRICS[10] and so on.

　　In addition, in the field of dose-response research, scholars have also applied many complex statistical methods, such as combining multiple comparisons and modeling techniques[11], adaptive dose-range design[12], optimal design of dose-response curve[13], Bayesian group sequential design[14] and so on. Therefore, based on previous explorations, this project strives to apply the above statistical methods to the study of the tumor chemosensitive assay in vitro of HB, hoping to make contributions to clinical drug guidance and improve and innovate the small sample statistical inference methods.