A comparison of machine learning algorithms and covariate balance measures for propensity score matching and weighting

Propensity score matching (PSM) and propensity score weighting (PSW) are popular tools to estimate causal effects in observational studies. We address two open issues: how to estimate propensity scores and assess covariate balance. Using simulations, we compare the performance of PSM and PSW based on logistic regression and machine learning algorithms (CART; Bagging; Boosting; Random Forest; Neural Networks; naive Bayes). Additionally, we consider several measures of covariate balance (Absolute Standardized Average Mean (ASAM) with and without interactions; measures based on the quantile‐quantile plots; ratio between variances of propensity scores; area under the curve (AUC)) and assess their ability in predicting the bias of PSM and PSW estimators. We also investigate the importance of tuning of machine learning parameters in the context of propensity score methods. Two simulation designs are employed. In the first, the generating processes are inspired to birth register data used to assess the effect of labor induction on the occurrence of caesarean section. The second exploits more general generating mechanisms. Overall, among the different techniques, random forests performed the best, especially in PSW. Logistic regression and neural networks also showed an excellent performance similar to that of random forests. As for covariate balance, the simplest and commonly used metric, the ASAM, showed a strong correlation with the bias of causal effects estimators. Our findings suggest that researchers should aim at obtaining an ASAM lower than 10% for as many variables as possible. In the empirical study we found that labor induction had a small and not statistically significant impact on caesarean section.     

On a class of non-linear transformation cure rate models

In this paper, we propose a generalization of the mixture (binary) cure rate model, motivated by the existence of a zero-modified (inflation or deflation) distribution, on the initial number of causes, under a competing cause scenario. This non-linear transformation cure rate model is in the same form of models studied in the past; however, following our approach, we are able to give a realistic interpretation to a specific class of proper transformation functions, for the cure rate modeling. The estimation of the parameters is then carried out using the maximum likelihood method along with a profile approach. A simulation study examines the accuracy of the proposed estimation method and the model discrimination based on the likelihood ratio test. For illustrative purposes, analysis of two real life data-sets, one on recidivism and another on cutaneous melanoma, is also carried out.     

A note on the interpretation of tree-based regression models

Tree-based models are a popular tool for predicting a response given a set of explanatory variables when the regression function is characterized by a certain degree of complexity. Sometimes, they are also used to identify important variables and for variable selection. We show that if the generating model contains chains of direct and indirect effects, then the typical variable importance measures suggest selecting as important mainly the background variables, which have a strong indirect effect, disregarding the variables that directly influence the response. This is attributable mainly to the variable choice in the first steps of the algorithm selecting the splitting variable and to the greedy nature of such search. This pitfall could be relevant when using tree-based algorithms for understanding the underlying generating process, for population segmentation and for causal inference.     

Immunomodulation of tahneeq method in IL-12 and CD8+ T-Lymphocyte,an in-vivo study in neonatal rats

Stimulation of the neonatal immune system is quite important for the proliferation and differentiation of antigen-presenting cells (APCs) and T cells. Tahneeq is a traditional method to manually rub the palatal mucosa of newborn babies with premasticated Ajwa palm dates. The present study was to investigate the tahneeq effects on IL-12 expression of dendritic cells (DCs) and blood T lymphocytes expressing CD8⁺ in neonatal Wistar rats. The number of 90 healthy neonatal Wistar rats have randomly divided into three groups: control group received breastmilk only, treatment group (T1) receiving breast milk + mild-scratched intensity of tahneeq, and T2 group received breastmilk + strong-scratched intensity of tahneeq on the palatal and gingival mucosa immediately after birth.  Seven neonatal Wistar rats in all groups were then sacrificed in three hours after birth and days 1, 5, 7, 13, and 30 treatment. IL-12 expression in the palatal and gingival mucosa was determined using immunohistochemical staining, and blood CD8⁺ T-lymphocytes were quantified using a flow cytometer. One way ANOVA was used to analyze the percentage of IL-12 and CD8⁺ T-lymphocytes among neonatal Wistar rat groups.  The T1 and T2 newborn rat groups had significantly higher IL-12 expression than the control group (p<0.001). The increased IL-12 expression in T2 groups significantly increased (p<0.001) compared to the IL-12 expression in the T1 and control groups. The percentage of  CD8⁺ T lymphocytes in all neonatal rat groups increased on three hours after birth and day 30 treatment but remained constant on days 5 and 7 treatment and decreased on day 13 treatment. At 5, 13, and 30^th days treatment,  the percentage of CD8⁺ T lymphocytes in T1 and T2 neonatal rat groups was significantly higher (p<0.05) than that in the control group. In conclusion, the impact on systemic CD8⁺ T cells did not influence by the depth of the scratch. Both mild and strong tahneeq increased the systemic CD8⁺ T-lymphocytes in neonatal Wistar rats. The roles of anti-inflammatory cytokines and Treg cells should be further investigated to unravel those different results for the development of mucosal immunity in neonates.     

siRNA表达载体的构建及其表达

本研究用限制性内切酶消化质粒pCMV-tag-2B,除去巨细胞病毒(Cytomegalovirus,CMV)启动子核苷酸序列,剩下的核苷酸序列作为构建表达siRNA(Smallinterfering RNA,siRNA)载体的前体.依据文献提供的扩增H1 RNA启动子核苷酸序列的引物序列合成一对引物,以带有H1 RNA启动子序列的质粒DNA为模板扩增H1RNA启动子序列,插入前体,构建SiRNA的表达载体pCH1.另外将H1 RNA启动子插入pGEM-11fz相应位点,构建瞬时表达载体pGH1.依据EGFP的有效SiRNA抑制位点,合成两条分别为64bp的核苷酸链,通过体外退火,形成双链,然后插入已构建的两个表达载体.将这两个载体分别与表达EGFP蛋白的质粒pEGFP-N3共转染Bel-7402细胞,观察siRNA对EGFP的抑制效应.研究结果表明构建的载体有效表达了siRNA,这些载体可以用于与siRNA相关抗病毒治疗性试验研究.     

Using genetic algorithms to select most predictive protein features

Many important characteristics of proteins such as biochemical activity and subcellular localization present a challenge to machine-learning methods: it is often difficult to encode the appropriate input features at the residue level for the purpose of making a prediction for the entire protein. The problem is usually that the biophysics of the connection between a machine-learning method's input (sequence feature) and its output (observed phenomenon to be predicted) remains unknown; in other words, we may only know that a certain protein is an enzyme (output) without knowing which region may contain the active site residues (input). The goal then becomes to dissect a protein into a vast set of sequence-derived features and to correlate those features with the desired output. We introduce a framework that begins with a set of global sequence features and then vastly expands the feature space by generically encoding the coexistence of residue-based features. It is this combination of individual features, that is the step from the fractions of serine and buried (input space 20 + 2) to the fraction of buried serine (input space 20 * 2) that implicitly shifts the search space from global feature inputs to features that can capture very local evidence such as a the individual residues of a catalytic triad. The vast feature space created is explored by a genetic algorithm (GA) paired with neural networks and support vector machines. We find that the GA is critical for selecting combinations of features that are neither too general resulting in poor performance, nor too specific, leading to overtraining. The final framework manages to effectively sample a feature space that is far too large for exhaustive enumeration. We demonstrate the power of the concept by applying it to prediction of protein enzymatic activity.