Search

There is much discussion concerning ‘digital transformation’ in healthcare and the potential of artificial intelligence (AI) in healthcare systems. Yet it remains rare to find AI solutions deployed in routine healthcare settings. This is in part due to the numerous challenges inherent in delivering an AI project in a clinical environment. In this article, several UK healthcare professionals and academics reflect on the challenges they have faced in building AI solutions using routinely collected healthcare data.These personal reflections are summarised as 10 practical tips. In our experience, these are essential considerations for an AI healthcare project to succeed. They are organised into four phases: conceptualisation, data management, AI application and clinical deployment. There is a focus on conceptualisation, reflecting our view that initial set-up is vital to success. We hope that our personal experiences will provide useful insights to others looking to improve patient care through optimal data use.

Abstract Background Transcription factor (TF) networks play a key role in controlling the transfer of genetic information from gene to mRNA. Much progress has been made on understanding and reverse-engineering TF network topologies using a range of experimental and theoretical methodologies. Less work has focused on using these models to examine how TF networks respond to changes in the cellular environment. Methods In this paper, we have developed a simple, pragmatic methodology, TIGERi (Transcription-factor-activity Illustrator for Global Explanation of Regulatory interaction), to model the response of an inferred TF network to changes in cellular environment. The methodology was tested using publicly available data comparing gene expression profiles of a mouse p38α (Mapk14) knock-out line to the original wild-type. Results Using the model, we have examined changes in the TF network resulting from the presence or absence of p38α. A part of this network was confirmed by experimental work in the original paper. Additional relationships were identified by our analysis, for example between p38α and HNF3, and between p38α and SOX9, and these are strongly supported by published evidence. FXR and MYC were also discovered in our analysis as two novel links of p38α. To provide a computational methodology to the biomedical communities that has more user-friendly interface, we also developed a standalone GUI (graphical user interface) software for TIGERi and it is freely available at https://github.com/namshik/tigeri/ . Conclusions We therefore believe that our computational approach can identify new members of networks and new interactions between members that are supported by published data but have not been integrated into the existing network models. Moreover, ones who want to analyze their own data with TIGERi could use the software without any command line experience. This work could therefore accelerate researches in transcriptional gene regulation in higher eukaryotes.

We have now reached the genomics era within medicine; genomics is being used to personalise treatment, make diagnoses, prognoses, and predict adverse outcomes resulting from treatment with certain drugs. Genomic data is now abundant in healthcare, and the newly created profession of clinical bioinformaticians are responsible for its analysis. In the United Kingdom, clinical bioinformaticians are trained within a 3-year programme, integrating a work-based placement with a part-time Master's degree. As this profession is still developing, trainees can feel isolated from their peers whom are located in other hospitals and can find it difficult to gain the mentorship that they require to complete their training. Building strong networks or communities of practice (CoPs) and allowing sharing of knowledge and experiences is one solution to addressing this isolation. Within the Master's delivered at the University of Manchester, we have integrated group-centred problem-based learning (PBL) using real clinical case studies worked on during each course unit. This approach is combined with a flipped style of teaching providing access to online content in our Virtual Learning Environment before the course. The face-to-face teaching is used to focus on the application of the students' knowledge to clinical case studies. In this study, we conducted semistructured interviews with 8 students, spanning 3 cohorts of students. We evaluated the effectiveness of this style of teaching and whether it had contributed to the formation of CoPs between our students. Our findings demonstrated that this style of teaching was preferred by our students to a more traditional lecture-based format and that the problem-based learning approach enabled the formation of CoPs within these cohorts. These CoPs are valuable in the development of this new profession and assist with the production of new guidelines and policies that are helping to professionalise this new group of healthcare scientists.

ABSTRACT Objectives The information contained within medical data is often used to make new medical discoveries. However, the most common way to use such data has been to query the data to answer very specific questions. For example, does having diabetes cause some patients to experience falls? If researchers have good questions, then the data can provide good answers. But are there any other equally important questions that could be asked of the data that people haven’t yet thought to ask? We are exploring a new strategy that we have developed to look for unusual and interesting patterns about falls in the elderly at subgroups level to see the different risks associated with different groups. Some of these risks will be associated with questions that are already well-known, but some should point to new and important questions that have not yet been asked. This opens up a better opportunity to identify patients at risk of falls, helping guide policy so as to reduce falls. Approach We mapped patient records into a low dimensional space using the notions of semantic similarity (Resnik node-based) and machine learning (principal component analysis) to provide a good representation of the data. This representation was used for clustering and visualisation through the DBSCAN algorithm. To look for enrichment in the resultant clusters, we analysed each cluster separately and look at the sets of patients defined in these clusters. Then, classic data mining techniques were used in order to generate hypotheses. The associations found were then be tested using more traditional comorbidity measures such as relative risk (RR) and its confidence intervals. Results We demonstrated the methodology on 589,169 older adults from clinical practice research datalink (CPRD). We successfully identified six distinct subgroups of falls from the elderly population who are identified with different risks. Some of the associations found are well defined in the literature; for example, depression and musculoskeletal conditions are significantly associated with falls. However, a number of associations are not reported in the clinical literature. Such hypotheses need further exploration by epidemiologists. Conclusion Future work will focus on incorporating temporal dimension which might provide useful insights into missed opportunities detection and risk modelling and understanding of a disease. Last, this methodology holds promises for the study of other complex diseases using any source of data which are described using terms from taxonomies or ontologies.

ABSTRACT Objectives Seasonal allergies, hay fever and asthma affect approximately one in four people and the incidence is increasing. Whilst the causes are unknown, hypotheses propose associations with environmental changes, exposure to pollutants and decreased exposure to childhood infections. High resolution datasets on pollen count and pollution are available; however there is no equivalent for incidence of seasonal allergy symptoms. We planned a national citizen science project using smartphones to gather data from the general population on seasonal allergy symptoms, and where and when they occur. The resulting dataset will be linked with other publicly available data, enabling better understanding of allergy triggers. This is a joint project between the Royal Society of Biology, the British Society for Immunology, and The University of Manchester. Approach In spring 2015, two codesign workshops were held for members of the public with seasonal allergies and/or asthma (n=33). Guided by researchers, attendees used paper prototyping techniques to illustrate the functionalities of a mobile application. They also prioritised functions for inclusion within the app, discussed data sharing options and suggested material for the accompanying project website (www.britainbreathing.org). Following codesign workshops, designs and requirements were collated, refined and used to build the first version of the application in Android. Results Workshop feedback indicated that potential users prioritised simple, personal tracking. They also valued the ability to access information about symptom frequencies among other users locally. Support for academic research was high, although most wanted some control over data sharing. People were comfortable with GPS data being collected, provided it did not impinge on privacy. We agreed to make data openly available via an interactive widget on the project website. The resulting first version of the application enables personal symptom tracking and will be released in March 2016 via the Google Play store (free of charge). A national media campaign will drive recruitment, alongside inclusion in the European City of Science 2016 programme in Manchester. Emerging data on the incidence of allergy symptoms by location will be presented. Conclusion Citizen science can be more than simply crowdsourcing data. We demonstrated that paper prototyping was a feasible and useful technique for codesigning an application with members of the public. Furthermore, workshop feedback indicated a high level of support for citizen science, provided users gained simple, personalised feedback. Further research is required to determine how codesign processes influence subsequent participant recruitment and engagement in citizen science projects.

In animal-based biomedical research, both the sex and the age of the animals studied affect disease phenotypes by modifying their susceptibility, presentation and response to treatment. The accurate reporting of experimental methods and materials, including the sex and age of animals, is essential so that other researchers can build on the results of such studies. Here we use text mining to study 15,311 research papers in which mice were the focus of the study. We find that the percentage of papers reporting the sex and age of mice has increased over the past two decades: however, only about 50% of the papers published in 2014 reported these two variables. We also compared the quality of reporting in six preclinical research areas and found evidence for different levels of sex-bias in these areas: the strongest male-bias was observed in cardiovascular disease models and the strongest female-bias was found in infectious disease models. These results demonstrate the ability of text mining to contribute to the ongoing debate about the reproducibility of research, and confirm the need to continue efforts to improve the reporting of experimental methods and materials.

Computer-based resources are central to much, if not most, biological and medical research. However, while there is an ever expanding choice of bioinformatics resources to use, described within the biomedical literature, little work to date has provided an evaluation of the full range of availability or levels of usage of database and software resources. Here we use text mining to process the PubMed Central full-text corpus, identifying mentions of databases or software within the scientific literature. We provide an audit of the resources contained within the biomedical literature, and a comparison of their relative usage, both over time and between the sub-disciplines of bioinformatics, biology and medicine. We find that trends in resource usage differs between these domains. The bioinformatics literature emphasises novel resource development, while database and software usage within biology and medicine is more stable and conservative. Many resources are only mentioned in the bioinformatics literature, with a relatively small number making it out into general biology, and fewer still into the medical literature. In addition, many resources are seeing a steady decline in their usage (e.g., BLAST, SWISS-PROT), though some are instead seeing rapid growth (e.g., the GO, R). We find a striking imbalance in resource usage with the top 5% of resource names (133 names) accounting for 47% of total usage, and over 70% of resources extracted being only mentioned once each. While these results highlight the dynamic and creative nature of bioinformatics research they raise questions about software reuse, choice and the sharing of bioinformatics practice. Is it acceptable that so many resources are apparently never reused? Finally, our work is a step towards automated extraction of scientific method from text. We make the dataset generated by our study available under the CC0 license here: http://dx.doi.org/10.6084/m9.figshare.1281371.

There is a growing concern both inside and outside the scientific community over the lack of reproducibility of experiments. The depth and detail of reported methods are critical to the reproducibility of findings, but also for making it possible to compare and integrate data from different studies. In this study, we evaluated in detail the methods reporting in a comprehensive set of trypanosomiasis experiments that should enable valid reproduction, integration and comparison of research findings. We evaluated a subset of other parasitic (Leishmania, Toxoplasma, Plasmodium, Trichuris and Schistosoma) and non-parasitic (Mycobacterium) experimental infections in order to compare the quality of method reporting more generally. A systematic review using PubMed (2000-2012) of all publications describing gene expression in cells and animals infected with Trypanosoma spp was undertaken based on PRISMA guidelines; 23 papers were identified and included. We defined a checklist of essential parameters that should be reported and have scored the number of those parameters that are reported for each publication. Bibliometric parameters (impact factor, citations and h-index) were used to look for association between Journal and Author status and the quality of method reporting. Trichuriasis experiments achieved the highest scores and included the only paper to score 100% in all criteria. The mean of scores achieved by Trypanosoma articles through the checklist was 65.5% (range 32-90%). Bibliometric parameters were not correlated with the quality of method reporting (Spearman's rank correlation coefficient 0.05). Our results indicate that the quality of methods reporting in experimental parasitology is a cause for concern and it has not improved over time, despite there being evidence that most of the assessed parameters do influence the results. We propose that our set of parameters be used as guidelines to improve the quality of the reporting of experimental infection models as a pre-requisite for integrating and comparing sets of data.

CONTEXT:Cortisol is released in ultradian pulses. The biological relevance of the resulting fluctuating cortisol concentration has not been explored. OBJECTIVE:Determination of the biological consequences of ultradian cortisol pulsatility. DESIGN:A novel flow through cell culture system was developed to deliver ultradian pulsed or continuous cortisol to cells. The effects of cortisol dynamics on cell proliferation and survival, and on gene expression were determined. In addition, effects on glucocorticoid receptor (GR) expression levels and phosphorylation, as a potential mediator, were measured. RESULTS:Pulsatile cortisol caused a significant reduction in cell survival compared to continuous exposure of the same cumulative dose, due to increased apoptosis. Comprehensive analysis of the transcriptome response by microarray identified genes with a differential response to pulsatile versus continuous glucocorticoid delivery. These were confirmed with qRT-PCR. Several transcription factor binding sites were enriched in these differentially regulated target genes, including CCAAT-displacement protein (CDP). A CDP regulated reporter gene (MMTV-luc) was, as predicted, also differentially regulated by pulsatile compared to continuous cortisol delivery. Importantly there was no effect of cortisol delivery kinetics on either GR expression, or activation (GR phosphoSer(211)). CONCLUSIONS:Cortisol oscillations exert important effects on target cell gene expression, and phenotype. This is not due to differences in cumulative cortisol exposure, or either expression, or activation of the GR. This suggests a novel means to regulate GR function.

African trypanosomes are protozoan parasites that cause "sleeping sickness" in humans and a similar disease in livestock. Trypanosomes also infect laboratory mice and three major quantitative trait loci (QTL) that regulate survival time after infection with T. congolense have been identified in two independent crosses between susceptible A/J and BALB/c mice, and the resistant C57BL/6. These were designated Tir1, Tir2 and Tir3 for Trypanosoma infection response, and range in size from 0.9-12 cM.Mapping loci regulating survival time after T. congolense infection in an additional cross revealed that susceptible C3H/HeJ mice have alleles that reduce survival time after infection at Tir1 and Tir3 QTL, but not at Tir2. Next-generation resequencing of a 6.2 Mbp region of mouse chromosome 17, which includes Tir1, identified 1,632 common single nucleotide polymorphisms (SNP) including a probably damaging non-synonymous SNP in Pram1 (PML-RAR alpha-regulated adaptor molecule 1), which was the most plausible candidate QTL gene in Tir1. Genome-wide comparative genomic hybridisation identified 12 loci with copy number variants (CNV) that correlate with differential gene expression, including Cd244 (natural killer cell receptor 2B4), which lies close to the peak of Tir3c and has gene expression that correlates with CNV and phenotype, making it a strong candidate QTL gene at this locus.By systematically combining next-generation DNA capture and sequencing, array-based comparative genomic hybridisation (aCGH), gene expression data and SNP annotation we have developed a strategy that can generate a short list of polymorphisms in candidate QTL genes that can be functionally tested.

Mammals are able to rapidly produce red blood cells in response to stress. The molecular pathways used in this process are important in understanding responses to anaemia in multiple biological settings. Here we characterise the novel gene Claudin 13 (Cldn13), a member of the Claudin family of tight junction proteins using RNA expression, microarray and phylogenetic analysis. We present evidence that Cldn13 appears to be co-ordinately regulated as part of a stress induced erythropoiesis pathway and is a mouse-specific gene mainly expressed in tissues associated with haematopoietic function. CLDN13 phylogenetically groups with its genomic neighbour CLDN4, a conserved tight junction protein with a putative role in epithelial to mesenchymal transition, suggesting a recent duplication event. Mechanisms of mammalian stress erythropoiesis are of importance in anaemic responses and expression microarray analyses demonstrate that Cldn13 is the most abundant Claudin in spleen from mice infected with Trypanosoma congolense. In mice prone to anaemia (C57BL/6), its expression is reduced compared to strains which display a less severe anaemic response (A/J and BALB/c) and is differentially regulated in spleen during disease progression. Genes clustering with Cldn13 on microarrays are key regulators of erythropoiesis (Tal1, Trim10, E2f2), erythrocyte membrane proteins (Rhd and Gypa), associated with red cell volume (Tmcc2) and indirectly associated with erythropoietic pathways (Cdca8, Cdkn2d, Cenpk). Relationships between genes appearing co-ordinately regulated with Cldn13 post-infection suggest new insights into the molecular regulation and pathways involved in stress induced erythropoiesis and suggest a novel, previously unreported role for claudins in correct cell polarisation and protein partitioning prior to erythroblast enucleation.

Trypanosoma congolense are extracellular protozoan parasites of the blood stream of artiodactyls and are one of the main constraints on cattle production in Africa. In cattle, anaemia is the key feature of disease and persists after parasitaemia has declined to low or undetectable levels, but treatment to clear the parasites usually resolves the anaemia.The progress of anaemia after Trypanosoma congolense infection was followed in three mouse strains. Anaemia developed rapidly in all three strains until the peak of the first wave of parasitaemia. This was followed by a second phase, characterized by slower progress to severe anaemia in C57BL/6, by slow recovery in surviving A/J and a rapid recovery in BALB/c. There was no association between parasitaemia and severity of anaemia. Furthermore, functional T lymphocytes are not required for the induction of anaemia, since suppression of T cell activity with Cyclosporin A had neither an effect on the course of infection nor on anaemia. Expression of genes involved in erythropoiesis and iron metabolism was followed in spleen, liver and kidney tissues in the three strains of mice using microarrays. There was no evidence for a response to erythropoietin, consistent with anaemia of chronic disease, which is erythropoietin insensitive. However, the expression of transcription factors and genes involved in erythropoiesis and haemolysis did correlate with the expression of the inflammatory cytokines Il6 and Ifng.The innate immune response appears to be the major contributor to the inflammation associated with anaemia since suppression of T cells with CsA had no observable effect. Several transcription factors regulating haematopoiesis, Tal1, Gata1, Zfpm1 and Klf1 were expressed at consistently lower levels in C57BL/6 mice suggesting that these mice have a lower haematopoietic capacity and therefore less ability to recover from haemolysis induced anaemia after infection.