Multilocus association mapping using generalized ridge logistic regression

Enlarged Lumen Volume of Proximal Aortic Segment and Acute Type B Aortic Dissection: A Computer Fluid Dynamics Study of Ideal Aortic Models

Background

Recent study has revealed that enlarged diameters of the ascending aorta and proximal aortic arch enhance the probability of ATBAD. However, little is understood about the relation to ATBAD.

Objective

This study explored the differences in proximal aortic segment (PAS) morphology in patients with acute type B aortic dissection (ATBAD), and performed hemodynamic simulations to provide proof of principle.

Materials and Methods

The morphological characteristics of PAS in the ATBAD group (n = 163) and corresponding segment in the control group (n = 120) were retrospectively measured. The morphological parameters were analyzed using comprehensive statistical approaches. Ridge regression analysis was also performed to determine the association between independent variable and dependent variable. P < 0.01 was considered significant. Idealized aortic models were established based on variables of statistical significance, and hemodynamic simulations were performed to evaluate blood flow changes caused by morphology.

Results

Diameters at landmarks of PAS were significantly larger in the ATBAD group. The lumen volume (V_PAS) of PAS in the ATBAD group was significantly enlarged than that of the control group (124,659.07 ± 34,089.27 mm³ vs 89,796.65 ± 30,334.40 mm³; P < 0.001). Furthermore, the V_PAS was positively correlated to diameters. As the V_PAS increased, the fluid kinetic energy in PAS enhanced linearly, and time-averaged wall shear stress and oscillatory shear index at the distal area of the left subclavian artery increased significantly.

Conclusion

In the ATBAD group, the enlarged V_PAS and increased diameters of PAS are positively correlated. Meanwhile, the enlarged V_PAS leads to more aggressive hemodynamic parameters at the distal area of the left subclavian artery, which may create a contributory condition for ATBAD.

Comorbid insomnia among breast cancer survivors and its prediction using machine learning: a nationwide study in Japan

OBJECTIVE

Insomnia is an increasingly recognized major symptom of breast cancer which can seriously disrupt the quality of life during and many years after treatment. Sleep problems have also been linked with survival in women with breast cancer. The aims of this study were to estimate the prevalence of insomnia in breast cancers survivors, clarify the clinical characteristics of their sleep difficulties and use machine learning techniques to explore clinical insights.

METHODS

Our analysis of data, obtained in a nationwide questionnaire survey of breast cancer survivors in Japan, revealed a prevalence of suspected insomnia of 37.5%. With the clinical data obtained, we then used machine learning algorithms to develop a classifier that predicts comorbid insomnia. The performance of the prediction model was evaluated using 8-fold cross-validation.

RESULTS

When using optimal hyperparameters, the L2 penalized logistic regression model and the XGBoost model provided predictive accuracy of 71.5 and 70.6% for the presence of suspected insomnia, with areas under the curve of 0.76 and 0.75, respectively. Population segments with high risk of insomnia were also extracted using the RuleFit algorithm. We found that cancer-related fatigue is a predictor of insomnia in breast cancer survivors.

CONCLUSIONS

The high prevalence of sleep problems and its link with mortality warrants routine screening. Our novel predictive model using a machine learning approach offers clinically important insights for the early detection of comorbid insomnia and intervention in breast cancer survivors.

DAPT score: predictive model of dual-antiplatelet therapy for acute cerebral infarction

OBJECTIVE

This article analyzes factors which affect the prognosis of acute cerebral infarction (ACI) patients receiving a course of antiplatelet therapy with aspirin and (or) clopidogrel for 14 days and proposes a simple grading scale to predict the clinical effectiveness of these drugs.

METHODS

We evaluated the association between ACI and risk factor (univariate analysis) on at day 14 post admission. Factors which potentially affected the 14-day prognosis of the patients were identified by logistic regression. A clinical grading scale, the DAPT score, was developed by weighing the independent predictors based on these factors.

RESULTS

It is revealed that the factors which affected 14 days prognosis univariate analysis included age ≥ 50 years (P = 0.007), diabetes (P = 0.017), hypertension (P ≤ 0.001), hyperhomocysteinemia (P = 0.001), and ipsilateral carotid artery stenosis ≥ 50% (P = 0.019). Logistic regression also revealed that the factors which affected 14 days prognosis included age ≥ 50 years (P = 0.007), hypertension (P ≤ 0.001), hyperhomocysteinemia (P = 0.001), and ipsilateral carotid artery stenosis ≥ 50% (P = 0.014).The assigned values of age ≥ 50 years, Grade 1 hypertension, Grade 2 hypertension, Grade 3 hypertension, hyperhomocysteinemia, and ipsilateral carotid artery stenosis ≥ 50% were 1, 1, 2, 3, 1, and 1, respectively. We named this score (DAPT score) and it ranged between 0 and 6. Using 3 as a cutoff, the sensitivity was 90.6% and the specificity was 63.3%.

CONCLUSIONS

The DAPT Score might be useful to for identifying with ACI who are suitable to receive aspirin combined with clopidogrel. Future large-scale, multi-center prospective studies are necessary.

Maximal Segmental Score Method for Localizing Recessive Disease Variants Based on Sequence Data

Background

Due to the affordability of whole-genome sequencing, the genetic association design can now address rare diseases. However, some common statistical association methods only consider homozygosity mapping and need several criteria, such as sliding windows of a given size and statistical significance threshold setting, such as P-value < 0.05 to achieve good power in rare disease association detection.

Methods

Our region-specific method, called expanded maximal segmental score (eMSS), converts p-values into continuous scores based on the maximal segmental score (MSS) (Lin et al., 2014) for detecting disease-associated segments. Our eMSS considers the whole genome sequence data, not only regions of homozygosity in candidate genes. Unlike sliding window methods of a given size, eMSS does not need predetermined parameters, such as window size or minimum or maximum number of SNPs in a segment. The performance of eMSS was evaluated by simulations and real data analysis for autosomal recessive diseases multiple intestinal atresia (MIA) and osteogenesis imperfecta (OI), where the number of cases is extremely small. For the real data, the results by eMSS were compared with a state-of-the-art method, HDR-del (Imai et al., 2016).

Results

Our simulation results show that eMSS had higher power as the number of non-causal haplotype blocks decreased. The type I error for eMSS under different scenarios was well controlled, p < 0.05. For our observed data, the bone morphogenetic protein 1 (BMP1) gene on chromosome 8, the Violaxanthin de-epoxidase-related chloroplast (VDR) gene on chromosome 12 associated with OI, and the tetratricopeptide repeat domain 7A (TTC7A) gene on chromosome 2 associated with MIA have previously been identified as harboring the relevant pathogenic mutations.

Conclusions

When compared to HDR-del, our eMSS is powerful in analyzing even small numbers of recessive cases, and the results show that the method can further reduce numbers of candidate variants to a very small set of susceptibility pathogenic variants underlying OI and MIA. When we conduct whole-genome sequence analysis, eMSS used 3/5 the computation time of HDR-del. Without additional parameters needing to be set in the segment detection, the computational burden for eMSS is lower compared with that in other region-specific approaches.

Supporting Regularized Logistic Regression Privately and Efficiently

As one of the most popular statistical and machine learning models, logistic regression with regularization has found wide adoption in biomedicine, social sciences, information technology, and so on. These domains often involve data of human subjects that are contingent upon strict privacy regulations. Concerns over data privacy make it increasingly difficult to coordinate and conduct large-scale collaborative studies, which typically rely on cross-institution data sharing and joint analysis. Our work here focuses on safeguarding regularized logistic regression, a widely-used statistical model while at the same time has not been investigated from a data security and privacy perspective. We consider a common use scenario of multi-institution collaborative studies, such as in the form of research consortia or networks as widely seen in genetics, epidemiology, social sciences, etc. To make our privacy-enhancing solution practical, we demonstrate a non-conventional and computationally efficient method leveraging distributing computing and strong cryptography to provide comprehensive protection over individual-level and summary data. Extensive empirical evaluations on several studies validate the privacy guarantee, efficiency and scalability of our proposal. We also discuss the practical implications of our solution for large-scale studies and applications from various disciplines, including genetic and biomedical studies, smart grid, network analysis, etc.

Using maximal segmental score in genome-wide association studies

Genome-wide association studies (GWAS) have become the method of choice for identifying disease susceptibility genes in common disease genetics research. Despite successes in these studies, much of the heritability remains unexplained due to lack of power and low resolution. High-density genotyping arrays can now screen more than 5 million genetic markers. As a result, multiple comparison has become an important issue especially in the era of next-generation sequencing. We propose to use a two-stage maximal segmental score procedure (MSS) which uses region-specific empirical P-values to identify genomic segments most likely harboring the disease gene. We develop scoring systems based on Fisher's P-value combining method to convert locus-specific significance levels into region-specific scores. Through simulations, our result indicated that MSS increased the power to detect genetic association as compared with conventional methods provided type I error was at 5%. We demonstrated the application of MSS on a publicly available case-control dataset of Parkinson's disease and replicated the findings in the literature. MSS provides an efficient exploratory tool for high-density association data in the current era of next-generation sequencing. R source codes to implement the MSS procedure are freely available at http://www.csjfann.ibms.sinica.edu.tw/EAG/program/programlist.htm.