MacLeod, Malcolm

Wegener, Gregers

Bannach-Brown, Alexandra

2019-10-14T10:28:56Z

2019-10-14T10:28:56Z

2019-11-25

Major Depressive Disorder (MDD) is the leading source of disability globally. Treatment-resistance among patients is common and even effective pharmacological therapies have a delayed effect on symptom relief. Better understanding of the mechanisms underlying depression and the search for potential effective and novel therapeutic targets are high research and healthcare priorities. Animal models are commonly used to mimic aspects of the phenotype of the human disorder to characterise candidate antidepressant agents. Despite these tools, no new pharmacological interventions have been discovered in the last decade and no reliable biomarkers have been identified for clinical use. Systematically reviewing the literature on animal models of depression may provide an overview of our current understanding of the underlying biological mechanisms and why no new therapies have been effectively translated to clinic. This field of research is large, and over 70,000 potentially relevant articles were identified in 2016. Therefore systematically reviewing this literature presents challenges for human resources. To combat these challenges, the following contributions to the field have been made: (1) the novel application of machine learning techniques to identify errors in human systematic review citation screening; and (2), the novel application of regular expression dictionaries to large corpuses of preclinical animal literature to help cluster publications into the disease model investigated and drug intervention tested. These tools have been applied for systematic review and meta-analysis methodology to the field of animal models of depression. All literature on animal models of depression has been systematically identified using searches carried out in PubMed and EMBASE in May 2016. This literature has been screened with the help of machine learning classification algorithms, based on a random set of dual human screened records (5749 records). This achieved a sensitivity of 98.7% and a specificity of 86% as assessed on in an independent validation dataset. Machine learning has been used to identify human screening errors in the set of documents used to train the algorithm. Correction of these errors with further human intervention, sees an improvement in specificity to 88.3%. These algorithms allow irrelevant documents to be automatically removed, reducing the corpus to 18,407 articles that highly likely to be relevant to the research area of animal models of depression. Custom-made regular expression dictionaries of (1) techniques to induce depressive-like phenotypes in animals, and (2) known antidepressants have been curated. The text-mining dictionaries for anti-depressant drugs and commonly used methods of model induction have been applied to categorise and visualise this large corpus of records to allow prioritisation of sub-topics of depression for further in depth systematic review and meta-analyses. These machine-assisted tools for systematic review methodology are available free to use, online. Systematic review and meta-analysis has been conducted on two sub-topics of the literature on animal models of depression. Firstly, the literature on the effects of ketamine as an anti-depressant in animal models of depression has been summarised with systematic review techniques and the effects of ketamine on depressive-like behaviour in the forced swim test, has been pooled using meta-analysis. The timing of administration of ketamine relative to the outcome assessment was significantly associated with decreases in effect size. This meta-analysis revealed no statistically significant heterogeneity between the studies. Secondly, the literature on use of gut microbial altering interventions to induce and treat depressive-like phenotypes in animal models of depression has been summarised and their effects have been pooled across studies using meta-analysis. The systematic review and meta-analysis of microbiota interventions identified a broad range of outcomes investigated in the primary literature and several probiotic treatments to reduce depressive-like behaviour were investigate gaps in the literature. Finally, a primary hypothesis-confirming animal experiment, where measures to reduce the risk of bias have been implemented was carried out to investigate the effects of prebiotics on depressive- and anxiety-like behaviour in a genetic animal model of depression, the Flinders Sensitive Line (FSL) rats. Online tools have been developed to provide an overview of animal models of depression and anti-depressant drugs investigated in the literature, using systematic review methodology and automation tools. This thesis reports meta-analyses on two sub-topics within animal models of depression; the effect of microbiota interventions, and the effects of ketamine; along with a primary animal experiment to test the effects of prebiotics on depressive-like behaviour in a genetic rodent model of depression.

http://hdl.handle.net/1842/36223

en

The University of Edinburgh

Bahor, Z., Liao, J., Macleod, M.R., Bannach-Brown, A., McCann, S.K., Wever, K.E., Thomas, J., Ottavi, T., Howells, D.W., Rice, A. and Ananiadou, S., 2017. Risk of bias reporting in the recent animal focal cerebral ischaemia literature. Clinical Science, 131(20), pp.2525-2532.

Bannach-Brown, A., Liao, J., Wegener, G., Macleod, M. R., 2016. Understanding in vivo modelling of depression: a systematic review protocol. CAMARADES repository of protocols. Retrieved from: https://drive.google.com/file/d/0BxckMffc78BYLWM2QUpBY3l1Q1k/view

Bannach-Brown, A., Thomas, J., Przybyła, P., Liao, J., (2016). “Protocol for Error Analysis: Machine learning and text mining solutions for systematic reviews of animal models of depression”. Published on CAMARADES Website. www.CAMARADES.info. Direct Access: https://drive.google.com/file/d/0BxckMffc78BYTm0tUzJJZkc1alk/view

Bannach-Brown, A., Przybyła, P., Thomas, J., Rice, A.S., Ananiadou, S., Liao, J. and Macleod, M.R., 2018. The use of text-mining and machine learning algorithms in systematic reviews: reducing workload in preclinical biomedical sciences and reducing human screening error. bioRxiv, p.255760.

Wang, Q., Liao, J., Hair, K., Bannach-Brown, A., Bahor, Z., Currie, G.L., McCann, S.K., Howells, D.W., Sena, E.S. and Macleod, M.R., 2018. Estimating the statistical performance of different approaches to meta-analysis of data from animal studies in identifying the impact of aspects of study design. bioRxiv, p.256776.

systematic review

Major Depressive Disorder

depression

meta-analysis

microbiome-targeting interventions

Understanding in vivo modelling of depression

Thesis or Dissertation

Doctoral

PhD Doctor of Philosophy