Development and validation of tools for systematic review and meta-analyses of complex, biological data sets

This PhD project will make a specific and unique contribution to developing, validating and operationalising "living" systematic review (SR) and meta-analysis analysis of complex biological datasets as part of the SLIM (Systemic Living Information Machine) programme; a CAMARADES (Collaborative Approach to Meta-Analysis and Review of Animal Data from Experimental Studies www.dcn.ed.ac.uk/camarades ) initiative.
SRs are vital research tools and are important for evidence-based decision making however, in preclinical research, the use of SR and meta-analysis (MA) are relatively novel. The methods for the conduct of SRs are well developed but they are resource intensive (Tricco et al., 2008) and the time they take to complete weakens their usefulness as they are often out of date once they have been published (Shojania et al., 2007). This problem is being exacerbated by the exponentially increasing number of publications in databases (Bastian et al., 2010). To address the complexities and challenges of research involving animals, CAMARADES have adapted the SR and MA techniques for use in the laboratory research arena to provide empirical evidence from which decisions can be made thereby reducing waste of research investment (Macleod et al., 2014).
Living SRs (LSR) are SRs that are continually updated, incorporating relevant new evidence as it becomes available (Elliott et al., 2017). Allowing humans and machines to interact and operate in mutually supportive ways to save time and improve accuracy. LSRs will be of great utility in fields of research which are moving quickly as they will increase efficiency and sustainability of SRs. This project will address a remaining challenge to implementation of LSR for in vivo biological data; outcome data extraction. Compared to the clinical data used in systematic review, basic science data is presented using a vast range and often idiosyncratic formats, therefore, a machine assisted approach coupled to crowd-based methods are the most feasible solution.
The aims of this project are three-fold: (i)To update and refine the bespoke machine-assisted data extraction toolkit specifically for machine assisted outcome data extraction in an in vivo pain neurobiology SR setting. (ii)To explore and prepare crowd-based approaches for toolkit deployment in LRS (iii) To test the toolkit in controlled trials, conducted using the SyRF platform in the context of exemplar automated SRs in three key areas of need in pain neurobiology: (i) Existing data: Determine normative values of lab rodent sensory thresholds to thermal, mechanical and cold stimuli. Evoked limb withdrawal to noxious heat, cold and mechanical stimuli is a ubiquitous and long standing method for determining sensory responses in rodents. (ii) Emerging data: measuring complex ethologically relevant pain behaviours in rodents and determining how these are perturbed by spontaneous pain is an emerging and rapidly growing area of pain research. As this field continues to expand it will benefit from an LSR being instituted at an early stage. (iii) Future opportunities: Many areas of neuroscience, including pain, now use video recording of a range of animal behaviours. This represents an opportunity to develop machine learning based automated analysis of digital files which could potentially be introduced into LSRs.
The development and incorporation of machine-assisted data extraction and MAs tools in open-source, online SR software would be beneficial and necessary to facilitate the production of accurate and timely evidence synthesis to improve decision making as well as contributing to the future functionality and success of LSRs. These powerful tools will assist in making sense of an exponentially growing body of data which will ultimately, make what was previously thought of as unattainable, attainable