Search

Plasma small RNAs have been recently explored as biomarkers in Huntington’s disease (HD). We performed an exploratory study on nine HD patients, eight healthy subjects (HS), and five psychiatric patients (PP; to control for iatrogenic confounder effects) through an Affymetrix-Gene-Chip-miRNA-Array. We validated the results in an independent population of 23 HD, 15 pre-HD, 24 PP, 28 Alzheimer’s disease (AD) patients (to control the disease-specificity) and 22 HS through real-time PCR. The microarray results showed higher levels of U13 small nucleolar RNA (SNORD13) in HD patients than controls (fold change 1.54, p = 0.003 HD vs. HS, and 1.44, p = 0.0026 HD vs. PP). In the validation population, a significant increase emerged with respect to both pre-HD and the control groups (p < 0.0001). SNORD13 correlated with the status of the mutant huntingtin carrier (r = 0.73; p < 0.001) and the disease duration (r = 0.59; p = 0.003). The receiver operating characteristic (ROC) curve analysis showed the high accuracy of SNORD13 in discriminating HD patients from other groups (AUC = 0.963). An interactome and pathway analysis on SNORD13 revealed enrichments for factors relevant to HD pathogenesis. We report the unprecedented finding of a potential disease-specific role of SNORD13 in HD. It seems to peripherally report a ‘tipping point’ in the pathogenic cascade at the neuronal level.

Machine learning has become an increasingly ubiquitous technology, as big data continues to inform and influence everyday life and decision-making. Currently in healthcare, as well as in most other industries, the two most prevalent machine learning paradigms are supervised learning and transfer learning. Both practices rely on large-scale, manually annotated datasets to train increasingly complex models. However, the requirement of data to be manually labeled leaves an excess of unused, unlabeled data available in both public and private data repositories. Self-supervised learning (SSL) is a growing area of machine learning that has the ability to take advantage of unlabeled data. Contrary to other machine learning paradigms, SSL algorithms create artificial supervisory signals from unlabeled data and pretrain algorithms on these signals. The aim of this review is two-fold: firstly, we provide a formal definition of SSL, divide SSL algorithms into their four unique subsets, and review the state-of-the-art published in each of those subsets between the years of 2014-2020. Second, this work surveys recent SSL algorithms published in healthcare, in order to provide medical experts with a clearer picture of how they can integrate SSL into their research, with the objective of leveraging unlabeled data.

Genome-wide association studies have identified more than 200 multiple sclerosis (MS)-associated loci across the human genome over the last decade, suggesting complexity in the disease etiology. This complexity poses at least two challenges: the definition of an etiological model including the impact of nongenetic factors, and the clinical translation of genomic data that may be drivers for new druggable targets. We reviewed studies dealing with single genes of interest, to understand how MS-associated single nucleotide polymorphism (SNP) variants affect the expression and the function of those genes. We then surveyed studies on the bioinformatic reworking of genome-wide association studies (GWAS) data, with aggregate analyses of many GWAS loci, each contributing with a small effect to the overall disease predisposition. These investigations uncovered new information, especially when combined with nongenetic factors having possible roles in the disease etiology. In this context, the interactome approach, defined as “modules of genes whose products are known to physically interact with environmental or human factors with plausible relevance for MS pathogenesis”, will be reported in detail. For a future perspective, a polygenic risk score, defined as a cumulative risk derived from aggregating the contributions of many DNA variants associated with a complex trait, may be integrated with data on environmental factors affecting the disease risk or protection.

// Evan L. Busch 1, 2, 3 , Jason L. Hornick 4 , Renato Umeton 5 , Adem Albayrak 5 , Neal I. Lindeman 4 , Laura E. MacConaill 4 , Elizabeth P. Garcia 4 , Matthew Ducar 4 and Timothy R. Rebbeck 2, 3 1 Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA 02115, USA 2 Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA 02115, USA 3 Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA 02215, USA 4 Department of Pathology, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA 02115, USA 5 Department of Informatics, Dana-Farber Cancer Institute, Boston, MA 02215, USA Correspondence to: Evan L. Busch, email: nhebu@channing.harvard.edu Keywords: metastasis, carcinomas, epithelial-mesenchymal transition, mutation, detachment Received: June 27, 2017 Accepted: September 01, 2017 Published: September 21, 2017 ABSTRACT Metastases are involved in most cancer deaths. Evidence has suggested that cancer cell detachment from primary tumors might occur largely via the mechanism of epithelial-mesenchymal transition (EMT) activated by epigenetic events, but data addressing other possible triggers of detachment, particularly genetic mutations, have been limited. Using the Profile study of cancer genomics at Dana-Farber Cancer Institute, we examined somatic mutations in the EMT genes CDH1 in 5,106 primary carcinomas and CTNNB1 in 7,578 primary carcinomas across 13 anatomic sites: urinary bladder, breast, colon/rectum, endometrium, esophagus, kidney, lung, ovary, pancreas, prostate, skin (non-melanoma), stomach, and thyroid. For each gene and anatomic site, we calculated the prevalence of primary carcinomas with at least one mutation. Across all anatomic sites, 4% of carcinomas had at least one CDH1 mutation and 4% of carcinomas had at least one CTNNB1 mutation. By anatomic site, the observed prevalence of carcinomas with at least one mutation was less than 5% at 10 sites for CDH1 and 12 sites for CTNNB1 . Tumor stage data were available for a subset of breast, colorectal, lung, and prostate tumors. Among patients from this subset who were diagnosed with regional or distant disease, only 4% had a CDH1 mutation and 1% had a CTNNB1 mutation in the primary tumor. The low mutation prevalences, especially among those with diagnoses of regional or distant disease, suggest that somatic mutations in CDH1 and CTNNB1 are unlikely to explain a substantial proportion of cancer cell detachment from primary carcinomas originating at most anatomic sites.

Several evidences emphasize B-cell pathogenic roles in multiple sclerosis (MS). We performed transcriptome analyses on peripheral B cells from therapy-free patients and age/sex-matched controls. Down-regulation of two transcripts (interferon response factor 1-IRF1, and C-X-C motif chemokine 10-CXCL10), belonging to the same pathway, was validated by RT-PCR in 26 patients and 21 controls. IRF1 and CXCL10 transcripts share potential seeding sequences for hsa-miR-424, that resulted up-regulated in MS patients. We confirmed this interaction and its functional effect by transfection experiments. Consistent findings indicate down-regulation of IRF1/CXCL10 axis, that may plausibly contribute to a pro-survival status of B cells in MS.

Changes of intestinal permeability (IP) have been extensively investigated in inflammatory bowel diseases (IBD) and celiac disease (CD), underpinned by a known unbalance between microbiota, IP and immune responses in the gut. Recently the influence of IP on brain function has greatly been appreciated. Previous works showed an increased IP that preceded experimental autoimmune encephalomyelitis development and worsened during disease with disruption of TJ. Moreover, studying co-morbidity between Crohn's disease and MS, a report described increased IP in a minority of cases with MS. In a recent work we found that an alteration of IP is a relatively frequent event in relapsing-remitting MS, with a possible genetic influence on the determinants of IP changes (as inferable from data on twins); IP changes included a deficit of the active mechanism of absorption from intestinal lumen. The results led us to hypothesize that gut may contribute to the development of MS, as suggested by another previous work of our group: a population of CD8+CD161high T cells, belonging to the mucosal-associated invariant T (MAIT) cells, a gut- and liver-homing subset, proved to be of relevance for MS pathogenesis. We eventually suggest future lines of research on IP in MS: studies on IP changes in patients under first-line oral drugs may result useful to improve their therapeutic index; correlating IP and microbiota changes, or IP and blood-brain barrier changes may help clarify disease pathogenesis; exploiting the IP data to disclose co-morbidities in MS, especially with CD and IBD, may be important for patient care.

There is no treatment for the myelin loss in multiple sclerosis, ultimately resulting in the axonal degeneration that leads to the progressive phase of the disease. We established a multi-tiered platform for the sequential screening of drugs that could be repurposed as remyelinating agents. We screened a library of 2,000 compounds (mainly Food and Drug Administration (FDA)-approved compounds and natural products) for cellular metabolic activity on mouse oligodendrocyte precursors (OPC), identifying 42 molecules with significant stimulating effects. We then characterized the effects of these compounds on OPC proliferation and differentiation in mouse glial cultures, and on myelination and remyelination in organotypic cultures. Three molecules, edaravone, 5-methyl-7-methoxyisoflavone and lovastatin, gave positive results in all screening tiers. We validated the results by retesting independent stocks of the compounds, analyzing their purity, and performing dose-response curves. To identify the chemical features that may be modified to enhance the compounds’ activity, we tested chemical analogs and identified, for edaravone, the functional groups that may be essential for its activity. Among the selected remyelinating candidates, edaravone appears to be of strong interest, also considering that this drug has been approved as a neuroprotective agent for acute ischemic stroke and amyotrophic lateral sclerosis in Japan.

As our knowledge about the etiology of multiple sclerosis (MS) increases, deterministic paradigms appear insufficient to describe the pathogenesis of the disease, and the impression is that stochastic phenomena (i.e. random events not necessarily resulting in disease in all individuals) may contribute to the development of MS. However, sources and mechanisms of stochastic behavior have not been investigated and there is no proposed framework to incorporate nondeterministic processes into disease biology. In this report, we will first describe analogies between physics of nonlinear systems and cell biology, showing how small-scale random perturbations can impact on large-scale phenomena, including cell function. We will then review growing and solid evidence showing that stochastic gene expression (or gene expression “noise”) can be a driver of phenotypic variation. Moreover, we will describe new methods that open unprecedented opportunities for the study of such phenomena in patients and the impact of this information on our understanding of MS course and therapy.

This work was funded by Italian Multiple Sclerosis Foundation grants (2007/R/17 and 2011/R/31) to MS. Though difficult, the study of gene-environment interactions in multifactorial diseases is crucial for interpreting the relevance of non-heritable factors and prevents from overlooking genetic associations with small but measurable effects. We propose a "candidate interactome" (i.e. a group of genes whose products are known to physically interact with environmental factors that may be relevant for disease pathogenesis) analysis of genome-wide association data in multiple sclerosis. We looked for statistical enrichment of associations among interactomes that, at the current state of knowledge, may be representative of gene-environment interactions of potential, uncertain or unlikely relevance for multiple sclerosis pathogenesis: Epstein-Barr virus, human immunodeficiency virus, hepatitis B virus, hepatitis C virus, cytomegalovirus, HHV8-Kaposi sarcoma, H1N1-influenza, JC virus, human innate immunity interactome for type I interferon, autoimmune regulator, vitamin D receptor, aryl hydrocarbon receptor and a panel of proteins targeted by 70 innate immune-modulating viral open reading frames from 30 viral species. Interactomes were either obtained from the literature or were manually curated. The P values of all single nucleotide polymorphism mapping to a given interactome were obtained from the last genome-wide association study of the International Multiple Sclerosis Genetics Consortium & the Wellcome Trust Case Control Consortium, 2. The interaction between genotype and Epstein Barr virus emerges as relevant for multiple sclerosis etiology. However, in line with recent data on the coexistence of common and unique strategies used by viruses to perturb the human molecular system, also other viruses have a similar potential, though probably less relevant in epidemiological terms. Publisher PDF

Heritable and nonheritable factors play a role in multiple sclerosis, but their effect size appears too small, explaining relatively little about disease etiology. Assuming that the factors that trigger the onset of the disease are, to some extent, also those that generate its remissions and relapses, we attempted to model the erratic behaviour of the disease course as observed on a dataset containing the time series of relapses and remissions of 70 patients free of disease-modifying therapies. We show that relapses and remissions follow exponential decaying distributions, excluding periodic recurrences and confirming that relapses manifest randomly in time. It is found that a mechanistic model with a random forcing describes in a satisfactory manner the occurrence of relapses and remissions, and the differences in the length of time spent in each one of the two states. This model may describe how interactions between “soft” etiologic factors occasionally reach the disease threshold thanks to comparably small external random perturbations. The model offers a new context to rethink key problems such as “missing heritability” and “hidden environmental structure” in the etiology of complex traits.

Background The information coming from biomedical ontologies and computational pathway models is expanding continuously: research communities keep this process up and their advances are generally shared by means of dedicated resources published on the web. In fact, such models are shared to provide the characterization of molecular processes, while biomedical ontologies detail a semantic context to the majority of those pathways. Recent advances in both fields pave the way for a scalable information integration based on aggregate knowledge repositories, but the lack of overall standard formats impedes this progress. Indeed, having different objectives and different abstraction levels, most of these resources "speak" different languages. Semantic web technologies are here explored as a means to address some of these problems. Methods Employing an extensible collection of interpreters, we developed OREMP (Ontology Reasoning Engine for Molecular Pathways), a system that abstracts the information from different resources and combines them together into a coherent ontology. Continuing this effort we present OREMPdb; once different pathways are fed into OREMP, species are linked to the external ontologies referred and to reactions in which they participate. Exploiting these links, the system builds species-sets, which encapsulate species that operate together. Composing all of the reactions together, the system computes all of the reaction paths from-and-to all of the species-sets. Results OREMP has been applied to the curated branch of BioModels (2011/04/15 release) which overall contains 326 models, 9244 reactions, and 5636 species. OREMPdb is the semantic dictionary created as a result, which is made of 7360 species-sets. For each one of these sets, OREMPdb links the original pathway and the link to the original paper where this information first appeared.

The Immune System is, together with Central Nervous System, one of the most important and complex unit of our organism. Despite great advances in recent years that shed light on its understanding and in the unraveling of key mechanisms behind its functions, there are still many areas of the Immune System that remain object of active research. The development of in-silico models, bridged with proper biological considerations, have recently improved the understanding of important complex systems [1,2]. In this paper, after introducing major role players and principal functions of the mammalian Immune System, we present two computational approaches to its modeling; i.e., two insilico Immune Systems. (i) A large-scale model, with a complexity of representation of 106 - 108 cells (e.g., APC, T, B and Plasma cells) and molecules (e.g., immunocomplexes), is here presented, and its evolution in time is shown to be mimicking an important region of a real immune response. (ii) Additionally, a viral infection model, stochastic and light-weight, is here presented as well: its seamless design from biological considerations, its modularity and its fast simulation times are strength points when compared to (i). Finally we report, with the intent of moving towards the virtual lymph note, a cost-benefits comparison among Immune System models presented in this paper.

In this paper we investigate plant photosynthesis and microbial fuel cells. We report the following: 1) we introduce and validate a novel multi-objective optimization algorithm, PMO2; 2) in photosynthesis we increase the yield of 135%, while in Geobacter sulfurreducens we determine the tradeoff for growth versus redox properties; 3) finally, we discuss Pareto-Front as an estimator of robust metabolic pathways.

We have studied the $\mathbf{C_3}$ photosynthetic carbon metabolism centering our investigation on the following four design principles. (1) Optimization of the photosynthetic rate by modifying the partitioning of resources between the different enzymes of the $\mathbf{C_3}$ photosynthetic carbon metabolism using a constant amount of protein-nitrogen. (2) Identify sensitive and less sensitive enzymes of the studied model. (3) Maximize photosynthetic productivity rate through the choice of robust enzyme concentrations using a new precise definition of robustness. (4) Modeling photosynthetic carbon metabolism as a multi-objective problem of two competing biological selection pressures: light-saturated photosynthetic rate versus total protein-nitrogen requirement. Using the designed single-objective optimization algorithms, PAO and A-CMA-ES, we have obtained an increase in photosynthetic productivity of the $\mathbf{135\%}$ from 15.486 $\mathbf{\mu mol~m^{-2}s^{-1}}$ to $\mathbf{36.382~\mu mol~m ^{-2}s^{-1}}$, and improving the previous best-found photosynthetic productivity value ($\mathbf{27.261}$ $\mathbf{\mu mol~m ^{-2}s^{-1}}$, $\mathbf{76\%}$ of enhancement). Optimized enzyme concentrations express a maximal local robustness ($\mathbf{100\%}$) and a high global robustness ($\mathbf{97.2\%}$), satisfactory properties for a possible ``in vitro'' manufacturing of the optimized pathway. Morris sensitivity analysis shows that 11 enzymes over 23 are high sensitive enzymes, i.e., the most influential enzymes of the carbon metabolism model. Finally, we have obtained the trade-off between the maximization of the leaf $\mathbf{CO_2}$ uptake rate and the minimization of the total protein-nitrogen concentration. This trade-off search has been carried out for the three $\mathbf{c_i}$ concentrations referring to the estimate of $\mathbf{CO_2}$ concentration in the atmosphere characteristic of 25 million years ago, nowadays and in 2100 a.C. Remarkably, the three Pareto frontiers identify the highest photosynthetic productivity rates together with the fewest protein-nitrogen usage.

The information coming from biomedical ontologies and runnable pathways is expanding continuously: research communities keep this process up and their advances are generally shared by means of dedicated resources published on the web. Having different objectives and different abstraction levels, most of these resources “speak” different languages. Employing an extensible collection of interpreters, we propose a system that abstracts the information from different resources and combines them together into a common meta-format. Preserving the resource independence, we provide an alignment service that can be used for multiple purposes. Two recent examples are: 1) The new web application Cytosolve uses an embedded version of this system to provide congruous parallel simulation of multiple models; 2) Using the BioModels.net database, a searchable dictionary of equivalent molecular reaction paths was built. Finally, the enriched knowledge can be exported in OWL and queried by semantically-enabled tools such as Protege. In this approach, we see a valuable tool to integrate and test information originating from different sources, while preserving the independence of the model curation process.

Machine Learning (ML) models are very effective in many learning tasks, due to the capability to extract meaningful information from large data sets. Nevertheless, there are learning problems that cannot be easily solved relying on pure data, e.g. scarce data or very complex functions to be approximated. Fortunately, in many contexts domain knowledge is explicitly available and can be used to train better ML models. This paper studies the improvements that can be obtained by integrating prior knowledge when dealing with a context-specific, non-trivial learning task, namely precision tuning of transprecision computing applications. The domain information is injected in the ML models in different ways: I) additional features, II) ad-hoc graph-based network topology, III) regularization schemes. The results clearly show that ML models exploiting problem-specific information outperform the data-driven ones, with an average improvement around 38%.