Testing non-additivity (interaction) in two-way ANOVA tables with no replication

Interaction effect: Are you doing the right thing?

How to correctly interpret interaction effects has been largely discussed in scientific literature. Nevertheless, misinterpretations are still frequently observed, and neuroscience is not exempt from this trend. We reviewed 645 papers published from 2019 to 2020 and found that, in the 93.2% of studies reporting a statistically significant interaction effect (N = 221), post-hoc pairwise comparisons were the designated method adopted to interpret its results. Given the widespread use of this approach, we aim to: (1) highlight its limitations and how it can lead to misinterpretations of the interaction effect; (2) discuss more effective and powerful ways to correctly interpret interaction effects, including both explorative and model selection procedures. The paper provides practical examples and freely accessible online materials to reproduce all analyses.

How Official Social Media Affected the Infodemic among Adults during the First Wave of COVID-19 in China

The COVID-19 pandemic has demonstrated that social media can impact society both positively (e.g., keeping citizens connected and informed) and negatively (e.g., the deliberate spreading of misinformation). This study aims to examine the underlying mechanisms of the relationship between official social media accounts and the infodemic, experienced during the first wave of COVID-19 in China. A theoretical model is proposed to examine how official social media accounts affected the infodemic during this period. In total, 1398 questionnaire responses were collected via WeChat and Tencent QQ, two leading Chinese social media platforms. Data analysis was conducted using Partial Lease Squares Structural Equation Modeling (PLS-SEM), moderation effect analysis, and mediation effect analysis. Results indicate that the Information Quality (IQ) of Official social media accounts (β = −0.294, p < 0.001) has a significant negative effect on the infodemic. Mediation effect analysis revealed that both social support (β = −0.333, 95% Boot CI (−0.388, −0.280)) and information cascades (β = −0.189, 95% Boot CI (−0.227, −0.151)) mediate the relationship between IQ and the infodemic. Moderation effect analysis shows that private social media usage (F = 85.637, p < 0.001) positively moderates the relationship between IQ and the infodemic, while health literacy has a small negative moderation effect on the relationship between IQ and the infodemic. Our findings show that, in the context of Chinese media, official social media accounts act as a major source of information for influencing the infodemic through increasing social support and reducing information cascades for citizens.

PRIMUS: Comprehensive proteomics of mouse intervertebral discs that inform novel biology and relevance to human disease modelling

•

Proteomics of healthy mouse IVDs differentiating compartments and spine levels.

•

NP cells feature vacuoles with lysosomal, transport and cell–cell communication functions.

•

Collagen XII, decorin and other ECM proteins contribute to function of the AF.

•

Distinct proteomics between lumbar and tail discs.

•

Mouse is a relevant model for human disc biology but care is needed in its use.

Mice are commonly used to study intervertebral disc (IVD) biology and related diseases such as IVD degeneration. Discs from both the lumbar and tail regions are used. However, little is known about compartmental characteristics in the different regions, nor their relevance to the human setting, where a functional IVD unit depends on a homeostatic proteome. Here, we address these major gaps through comprehensive proteomic profiling and in-depth analyses of 8-week-old healthy murine discs, followed by comparisons with human. Leveraging on a dataset of over 2,700 proteins from 31 proteomic profiles, we identified key molecular and cellular differences between disc compartments and spine levels, but not gender. The nucleus pulposus (NP) and annulus fibrosus (AF) compartments differ the most, both in matrisome and cellularity contents. Differences in the matrisome are consistent with the fibrous nature required for tensile strength in the AF and hydration property in the NP. Novel findings for the NP cells included an enrichment in cell junction proteins for cell–cell communication (Cdh2, Dsp and Gja1) and osmoregulation (Slc12a2 and Wnk1). In NP cells, we detected heterogeneity of vacuolar organelles; where about half have potential lysosomal function (Vamp3, Copb2, Lamp1/2, Lamtor1), some contain lipid droplets and others with undefined contents. The AF is enriched in proteins for the oxidative stress responses (Sod3 and Clu). Interestingly, mitochondrial proteins are elevated in the lumbar than tail IVDs that may reflect differences in metabolic requirement. Relative to the human, cellular and structural information are conserved for the AF. Even though the NP is more divergent between mouse and human, there are similarities at the level of cell biology. Further, common cross-species markers were identified for both NP (KRT8/19, CD109) and AF (COL12A1). Overall, mouse is a relevant model to study IVD biology, and an understanding of the limitation will facilitate research planning and data interpretation, maximizing the translation of research findings to human IVDs.

REMAXINT: a two-mode clustering-based method for statistical inference on two-way interaction

We present a novel method, REMAXINT, that captures the gist of two-way interaction in row by column (i.e., two-mode) data, with one observation per cell. REMAXINT is a probabilistic two-mode clustering model that yields two-mode partitions with maximal interaction between row and column clusters. For estimation of the parameters of REMAXINT, we maximize a conditional classification likelihood in which the random row (or column) main effects are conditioned out. For testing the null hypothesis of no interaction between row and column clusters, we propose a $$max-F$$ m a x - F test statistic and discuss its properties. We develop a Monte Carlo approach to obtain its sampling distribution under the null hypothesis. We evaluate the performance of the method through simulation studies. Specifically, for selected values of data size and (true) numbers of clusters, we obtain critical values of the $$max-F$$ m a x - F statistic, determine empirical Type I error rate of the proposed inferential procedure and study its power to reject the null hypothesis. Next, we show that the novel method is useful in a variety of applications by presenting two empirical case studies and end with some concluding remarks.

A Tumbling Magnetic Microrobot System for Biomedical Applications

A microrobot system comprising an untethered tumbling magnetic microrobot, a two-degree-of-freedom rotating permanent magnet, and an ultrasound imaging system has been developed for in vitro and in vivo biomedical applications. The microrobot tumbles end-over-end in a net forward motion due to applied magnetic torque from the rotating magnet. By turning the rotational axis of the magnet, two-dimensional directional control is possible and the microrobot was steered along various trajectories, including a circular path and P-shaped path. The microrobot is capable of moving over the unstructured terrain within a murine colon in in vitro, in situ, and in vivo conditions, as well as a porcine colon in ex vivo conditions. High-frequency ultrasound imaging allows for real-time determination of the microrobot's position while it is optically occluded by animal tissue. When coated with a fluorescein payload, the microrobot was shown to release the majority of the payload over a 1-h time period in phosphate-buffered saline. Cytotoxicity tests demonstrated that the microrobot's constituent materials, SU-8 and polydimethylsiloxane (PDMS), did not show a statistically significant difference in toxicity to murine fibroblasts from the negative control, even when the materials were doped with magnetic neodymium microparticles. The microrobot system's capabilities make it promising for targeted drug delivery and other in vivo biomedical applications.

Gender-specific differences in clinical and metabolic variables associated with NAFLD in a Mexican pediatric population

INTRODUCTION AND OBJECTIVES

Non-alcoholic fatty liver disease (NAFLD) is the most common liver disease in children and it is more prevalent in Hispanic males. The gender differences can be explained by body fat distribution, lifestyle, or sex hormone metabolism. We evaluated anthropometric and metabolic differences by gender in children with and without NAFLD.

METHODS

We included 194 participants (eutrophic, overweight, and individuals with obesity). The presence of NAFLD was determined using ultrasonography, and we evaluated the association between this disease with metabolic and anthropometric variables by gender.

RESULTS

The mean age was 10.64±2.54 years. The frequency of NAFLD in boys was 24.51% and in girls was 11.96% (OR=2.39; 95%CI=1.10-5.19; p=0.025). For girls, NAFLD was significantly associated with triglycerides (p=0.012), homeostatic model assessment of insulin resistance (HOMA-IR) (p=0.048), and the visceral adiposity index (VAI) (p=0.024). The variables related to NAFLD in a gender-specific manner were body mass index (BMI) (p=0.001), waist circumference (WC) (p<0.001), HDL cholesterol (p=0.021), alanine aminotransferase (ALT) (p<0.001), and aspartate aminotransferase (AST) (p=0.002).

CONCLUSIONS

In our study NAFLD is more frequent in boys, only ALT, and no other clinical or metabolic variables, were associated with NAFLD in these patients. HOMA-IR, VAI, triglyceride levels, and ALT were associated with NAFLD only in girls. The ALT cut-off points for the development of NAFLD in our study were 28.5U/L in females and 27.5U/L in males. Our findings showed that NAFLD should be intentionally screened in patients with obesity, particularly in boys.

Reversal of Chemoresistance in Ovarian Cancer by Co-Delivery of a P-Glycoprotein Inhibitor and Paclitaxel in a Liposomal Platform

The overexpression of permeability-glycoprotein (P-gp), an ABC transporter involved in the cellular exclusion of chemotherapeutic drugs, is a major factor in paclitaxel-resistant ovarian cancer. However, in clinical trials, co-administration of P-gp inhibitors and anticancer drugs has not resulted in the efficient reversal of drug resistance. To improve administration, we encapsulated the third-generation P-gp inhibitor tariquidar (XR-9576, XR), alone or in combination with paclitaxel (PCT) in liposomes (LP). After optimization, the liposomes demonstrated favorable physicochemical properties and the ability to reverse chemoresistance in experiments using chemosensitive/chemoresistant ovarian cancer cell line pairs. Analyzing publicly available datasets, we found that overexpression of P-gp in ovarian cancer is associated with a shorter progression-free and overall survival. In vitro, LP(XR) significantly increased the cellular retention of rhodamine 123, a P-gp substrate. LP(XR,PCT) synergistically inhibited cell viability, blocked proliferation, and caused G₂-M arrest in paclitaxel-resistant SKOV3-TR and HeyA8-MDR cell lines overexpressing P-gp. Holographic imaging cytometry revealed that LP(XR,PCT) treatment of SKOV3-TR cells induced almost complete mitotic arrest, whereas laser scanning cytometry showed that the treatment induced apoptosis. In proof-of-concept preclinical studies, LP(XR,PCT), when compared with LP(PCT), significantly reduced tumor weight (43.2% vs. 16.9%, P = 0.0007) and number of metastases (44.4% vs. 2.8%, P = 0.012) in mice bearing orthotopic HeyA8-MDR ovarian tumors. In the xenografts, LP(XR,PCT) efficiently induced apoptosis and impaired proliferation. Our findings suggest that co-delivery of a P-gp inhibitor and paclitaxel using a liposomal platform can sensitize paclitaxel-resistant ovarian cancer cells to paclitaxel. LP(XR,PCT) should be considered for clinical testing in patients with P-gp-overexpressing tumors. Mol Cancer Ther; 15(10); 2282-93. ©2016 AACR.

ADAM10 mediates trastuzumab resistance and is correlated with survival in HER2 positive breast cancer

Trastuzumab prolongs survival in HER2 positive breast cancer patients. However, resistance remains a challenge. We have previously shown that ADAM17 plays a key role in maintaining HER2 phosphorylation during trastuzumab treatment. Beside ADAM17, ADAM10 is the other well characterized ADAM protease responsible for HER ligand shedding. Therefore, we studied the role of ADAM10 in relation to trastuzumab treatment and resistance in HER2 positive breast cancer. ADAM10 expression was assessed in HER2 positive breast cancer cell lines and xenograft mice treated with trastuzumab. Trastuzumab treatment increased ADAM10 levels in HER2 positive breast cancer cells (p ≤ 0.001 in BT474; p ≤ 0.01 in SKBR3) and in vivo (p ≤ 0.0001) compared to control, correlating with a decrease in PKB phosphorylation. ADAM10 inhibition or knockdown enhanced trastuzumab response in naïve and trastuzumab resistant breast cancer cells. Trastuzumab monotherapy upregulated ADAM10 (p ≤ 0.05); and higher pre-treatment ADAM10 levels correlated with decreased clinical response (p ≤ 0.05) at day 21 in HER2 positive breast cancer patients undergoing a trastuzumab treatment window study. Higher ADAM10 levels correlated with poorer relapse-free survival (p ≤ 0.01) in a cohort of HER2 positive breast cancer patients. Our studies implicate a role of ADAM10 in acquired resistance to trastuzumab and establish ADAM10 as a therapeutic target and a potential biomarker for HER2 positive breast cancer patients.

Factor selection and structural identification in the interaction ANOVA model

When faced with categorical predictors and a continuous response, the objective of an analysis often consists of two tasks: finding which factors are important and determining which levels of the factors differ significantly from one another. Often times, these tasks are done separately using Analysis of Variance (ANOVA) followed by a post hoc hypothesis testing procedure such as Tukey's Honestly Significant Difference test. When interactions between factors are included in the model the collapsing of levels of a factor becomes a more difficult problem. When testing for differences between two levels of a factor, claiming no difference would refer not only to equality of main effects, but also to equality of each interaction involving those levels. This structure between the main effects and interactions in a model is similar to the idea of heredity used in regression models. This article introduces a new method for accomplishing both of the common analysis tasks simultaneously in an interaction model while also adhering to the heredity-type constraint on the model. An appropriate penalization is constructed that encourages levels of factors to collapse and entire factors to be set to zero. It is shown that the procedure has the oracle property implying that asymptotically it performs as well as if the exact structure were known beforehand. We also discuss the application to estimating interactions in the unreplicated case. Simulation studies show the procedure outperforms post hoc hypothesis testing procedures as well as similar methods that do not include a structural constraint. The method is also illustrated using a real data example.

Principal interactions analysis for repeated measures data: application to gene-gene and gene-environment interactions

Many existing cohorts with longitudinal data on environmental exposures, occupational history, lifestyle/ behavioral characteristics, and health outcomes have collected genetic data in recent years. In this paper, we consider the problem of modeling gene-gene and gene-environment interactions with repeated measures data on a quantitative trait. We review possibilities of using classical models proposed by Tukey (1949) and Mandel (1961) using the cell means of a two-way classification array for such data. Although these models are effective for detecting interactions in the presence of main effects, they fail miserably if the interaction structure is misspecified. We explore a more robust class of interaction models that are based on a singular value decomposition of the cell-means residual matrix after fitting the additive main effect terms. This class of additive main effects and multiplicative interaction models (Gollob, 1968) provide useful summaries for subject-specific and time-varying effects as represented in terms of their contribution to the leading eigenvalues of the interaction matrix. It also makes the interaction structure more amenable to geometric representation. We call this analysis 'principal interactions analysis'. While the paper primarily focuses on a cell-mean-based analysis of repeated measures outcome, we also introduce resampling-based methods that appropriately recognize the unbalanced and longitudinal nature of the data instead of reducing the response to cell means. We illustrate the proposed methods by using data from the Normative Aging Study, a longitudinal cohort study of Boston area veterans since 1963. We carry out simulation studies under an array of classical interaction models and common epistasis models to illustrate the properties of the principal interactions analysis procedure in comparison with the classical alternatives.

Temporal analysis of feeding patterns of Culex erraticus in central Alabama

Host blood meals in seven mosquito species previously shown to be infected with eastern equine encephalitis virus at a site in the Tuskegee National Forest in southcentral Alabama were investigated. Of 1374 blood meals derived from 88 different host species collected over 6 years from these seven mosquito species, 1099 were derived from Culex erraticus. Analysis of the temporal pattern of Cx. erraticus meals using a Runs test revealed that the patterns of feeding upon avian and mammalian hosts from March to September of each year were not randomly distributed over time. Similarly, meals taken from the three most commonly targeted host species (yellow-crowned night heron, great blue heron, and white-tailed deer) were not randomly distributed. A Tukey's two-way analysis of variance test demonstrated that although the temporal pattern of meals taken from avian hosts were consistent over the years, the patterns of meals taken from the individual host species were not consistent from year to year.