Multilevel Latent Differential Structural Equation Model with Short Time Series and Time-Varying Covariates: A Comparison of Frequentist and Bayesian Estimators.

Continuous-time modeling using differential equations is a promising technique to model change processes with longitudinal data. Among ways to fit this model, the Latent Differential Structural Equation Modeling (LDSEM) approach defines latent derivative variables within a structural equation modeling (SEM) framework, thereby allowing researchers to leverage advantages of the SEM framework for model building, estimation, inference, and comparison purposes. Still, a few issues remain unresolved, including performance of multilevel variations of the LDSEM under short time lengths (e.g., 14 time points), particularly when coupled multivariate processes and time-varying covariates are involved. Additionally, the possibility of using Bayesian estimation to facilitate the estimation of multilevel LDSEM (M-LDSEM) models with complex and higher-dimensional random effect structures has not been investigated. We present a series of Monte Carlo simulations to evaluate three possible approaches to fitting M-LDSEM, including: frequentist single-level and two-level robust estimators and Bayesian two-level estimator. Our findings suggested that the Bayesian approach outperformed other frequentist approaches. The effects of time-varying covariates are well recovered, and coupling parameters are the least biased especially using higher-order derivative information with the Bayesian estimator. Finally, an empirical example is provided to show the applicability of the approach.

From intention to behavior: Within- and between-person moderators of the relationship between intention and physical activity.

Intention is a proximal predictor of behavior in many theories of behavior change, but intentions to be physically active do not always translate to actual physical activity. Little research has examined intensive longitudinal changes in physical activity and corresponding within-person moderators needed to elucidate the mechanisms, hurdles, and facilitators of individuals' everyday physical activity behaviors. The present study set out to evaluate the possible moderators of the intention-physical activity relationship across within-person and between-person levels, including cross-level interactions. Data comprise the first intensive measurement burst (14 days) of the longitudinal prospective Healthy Aging in Industrial Environment (HAIE) study, with N = 1135 participants (N = 10,030 person-days), aged 18-65. Physical activity was operationalized as step counts measured objectively using Fitbit Charge 3/4 fitness monitor. Intention, barriers to physical activity, and social support for physical activity were measured daily via smartphone surveys. Stable characteristics, i.e., physical activity habit and exercise identity, were measured using an online questionnaire. A multilevel moderation regression model with Bayesian estimator was fitted. At the within-person level, the relation between intention and steps was weaker on days when barriers were more severe than usual for a given person (Estimate = -0.267; CI95 = [-0.340, -0.196]) and social support was below average for a given person (Est = 0.143; CI95 = [0.023, 0.262]). Additionally, the daily intention-behavior relationship was stronger for people with lower average severity of barriers (Est = -0.153; CI95 = [-0.268, -0.052]), higher exercise identity (Est = 0.300; CI95 = [0.047, 0.546]), men (Est = -1.294, CI95 = [-1.854, -0.707]), and older individuals (Est = 0.042, CI95 = [0.017, 0.064]). At the between-person level, only physical activity habit strengthened the intention-behavior link (Est = 0.794; CI95 = [0.090, 1.486]). Our results underscore the need to separate the between-person differences from the within-person fluctuations to better understand the individual dynamics in physical activity behaviors. Personalized interventions aimed at helping individuals translate intentions to actual physical activity could be tailored and become more intensive when there is a higher risk of intention-behavior gap on a given day for a specific individual (i.e., a day with more severe barriers and less social support), by increasing the dosage or deploying more precisely targeted intervention strategies and components. In addition, interventionists should take gender and age into account when tailoring everyday strategies to help individuals act on their intentions.

Patient-derived organoids as a preclinical platform for precision medicine in colorectal cancer.

Patient-derived organoids are being considered as models that can help guide personalized therapy through in vitro anticancer drug response evaluation. However, attempts to quantify in vitro drug responses in organoids and compare them with responses in matched patients remain inadequate. In this study, we investigated whether drug responses of organoids correlate with clinical responses of matched patients and disease progression of patients. Organoids were established from 54 patients with colorectal cancer who (except for one patient) did not receive any form of therapy before, and tumor organoids were assessed through whole-exome sequencing. For comparisons of in vitro drug responses in matched patients, we developed an 'organoid score' based on the variable anticancer treatment responses observed in organoids. Very interestingly, a higher organoid score was significantly correlated with a lower tumor regression rate after the standard-of-care treatment in matched patients. Additionally, we confirmed that patients with a higher organoid score (≥ 2.5) had poorer progression-free survival compared with those with a lower organoid score (< 2.5). Furthermore, to assess potential drug repurposing using an FDA-approved drug library, ten tumor organoids derived from patients with disease progression were applied to a simulation platform. Taken together, organoids and organoid scores can facilitate the prediction of anticancer therapy efficacy, and they can be used as a simulation model to determine the next therapeutic options through drug screening. Organoids will be an attractive platform to enable the implementation of personalized therapy for colorectal cancer patients.

Phenotype-based single cell sequencing identifies diverse genetic subclones in CD133 positive cancer stem cells.

Tumor heterogeneity is one of the ongoing huddles in the field of colon cancer therapy. It is evident that there are countless clones which exhibit different phenotypes and therefore, single cell analysis is inevitable. Cancer stem cells (CSCs) are rare cell population within tumor which is known to function in cancer metastasis and recurrence. Although there have been trials to prove intra-tumoral heterogeneity using single cell sequencing, that of CSCs has not been clearly elucidated. Here, we articulate the presence of heterogeneous subclones within CD133 positive cancer stem cells through single cell sequencing. As a proof of principle, we performed phenotype-based high-throughput laser isolation and single cell sequencing (PHLI-seq) of CD133 positive cells in a frozen tumor tissue obtained from a patient with colorectal cancer. The result proved that CD133 positive cells were shown to be heterogeneous both in copy number and mutational profiles. Single cancer stem cell specific mutations such as RNF144A, PAK2, PARP4, ADAM21, HYDIN, KRT38 and CELSR1 could be also detected in liver metastatic tumor of the same patient. Collectively, these data suggest that single cell analysis used to spot subclones with genetic variation within rare population, will lead to new strategies to tackle colon cancer metastasis.

Activation of WNT/ß-catenin signaling results in resistance to a dual PI3K/mTOR inhibitor in colorectal cancer cells harboring PIK3CA mutations.

PIK3CA is a frequently mutated gene in cancer, including about ~15 to 20% of colorectal cancers (CRC). PIK3CA mutations lead to activation of the PI3K/AKT/mTOR signaling pathway, which plays pivotal roles in tumorigenesis. Here, we investigated the mechanism of resistance of PIK3CA-mutant CRC cell lines to gedatolisib, a dual PI3K/mTOR inhibitor. Out of a panel of 29 CRC cell lines, we identified 7 harboring one or more PIK3CA mutations; of these, 5 and 2 were found to be sensitive and resistant to gedatolisib, respectively. Both of the gedatolisib-resistant cell lines expressed high levels of active glycogen synthase kinase 3-beta (GSK3ß) and harbored the same frameshift mutation (c.465_466insC; H155fs*) in TCF7, which encodes a positive transcriptional regulator of the WNT/ß-catenin signaling pathway. Inhibition of GSK3ß activity in gedatolisib-resistant cells by siRNA-mediated knockdown or treatment with a GSK3ß-specific inhibitor effectively reduced the activity of molecules downstream of mTOR and also decreased signaling through the WNT/ß-catenin pathway. Notably, GSK3ß inhibition rendered the resistant cell lines sensitive to gedatolisib cytotoxicity, both in vitro and in a mouse xenograft model. Taken together, these data demonstrate that aberrant regulation of WNT/ß-catenin signaling and active GSK3ß induced by the TCF7 frameshift mutation cause resistance to the dual PI3K/mTOR inhibitor gedatolisib. Cotreatment with GSK3ß inhibitors may be a strategy to overcome the resistance of PIK3CA- and TCF7-mutant CRC to PI3K/mTOR-targeted therapies.

Deep Learning in Medical Imaging: General Overview.

The artificial neural network (ANN)-a machine learning technique inspired by the human neuronal synapse system-was introduced in the 1950s. However, the ANN was previously limited in its ability to solve actual problems, due to the vanishing gradient and overfitting problems with training of deep architecture, lack of computing power, and primarily the absence of sufficient data to train the computer system. Interest in this concept has lately resurfaced, due to the availability of big data, enhanced computing power with the current graphics processing units, and novel algorithms to train the deep neural network. Recent studies on this technology suggest its potentially to perform better than humans in some visual and auditory recognition tasks, which may portend its applications in medicine and healthcare, especially in medical imaging, in the foreseeable future. This review article offers perspectives on the history, development, and applications of deep learning technology, particularly regarding its applications in medical imaging.