skip to content

Centre for Integrative Neuroscience Discovery

The project’s central aim is producing a robust and reproducible methodology for network analysis in neuroscience. We are building an accessible toolset fully integrated with existing workflows that provides more intuitive use possibilities for conducting graph analysis of biological data. The project will translate graph theoretical quantification and statistical evaluation into an open and reproducible workflow for network analyses of biological systems.

 

The codebase we have created has a robust series of classes and functions to generate workflows in the form of analysis pipelines. In addition to the workflow codebase much code has been generated to facilitate easy analysis of imaging data. This includes: 1) streamlined API calls to existing packages e.g. Tensorflow and fmriprep; 2) In-house efficient implementation of analysis and statistical methodologies such as Modularity Overlap and Spin distributions; 3) convenience wrappers for common function calls and analysis queries. Finally, the package accepts any python callable object allowing for completely generalized analysis formats to be implemented. When these become increasingly common they may be included as convenience functions by the community. The existing codebase has allowed us to already reproduce and validate existing work done in the Cambridge neuroscience community. An installable version of the package exists at www.github.com/Nick-Gale/NetworkAnalysis but has not been deployed in mainline Python package distribution.

 

Figure 1: Reproduction of the degree distribution analysis presented by Bethlehem et. al (2017). The graphpype workflow is able to show independently that the research reproduces while extending the plot ranges to reveal a broader story which can be used to prompt further research outside the scope of the original paper.

 

The project has now moved into the documentation and dissemination phase and we have generated a user-friendly readthedocs website https://nick-gale.github.io/graphpypeDocs/. This documentation will be continually updated with new tutorials and user-guides. The website follows a common style format, numpydoc and reStructuredText which allows for easy contribution by the community: a major goal of the project going forward. These documents will form the basis of workshops/tutorials to encourage usage of the package as well as community contribution. Further to this, we hope to publish a paper summarizing the codebase and its usefulness with JOSS and GigaScience being identified as target journals.

 

Future work includes: 1) finishing the user-guides and tutorial section of the documentation; 2) presenting the package in workshops; 3) continually updating the codebase and accompanying documentation; 4) publishing the work to date; 5) encouraging user adoption and contribution from the Cambridge neuroscience community.

 

Summary

1. A prototype recipe and pipeline format has been developed to enable shareable and

robustly reproducible neuroimaging analysis.

2. A code base with efficient native code and API linkage has been developed to generate

analysis pipelines and has been deployed to reproduce existing neuroimaging research.

3. The package is well-documented;members of the community can contribute easily to the package..

4. The project aims to move toward dissemination of the package  into the well-established Python packaging repositories

 

[1] Bethlehem, R. A., Romero-Garcia, R., Mak, E., Bullmore, E. T., & Baron-Cohen, S. (2017). Structural covariance networks in children with autism or ADHD. Cerebral Cortex, 27(8), 4267-4276.

About Us

The Centre for Integrative Neuroscience Discovery (CIND) brings together researchers working at the intersections of neurocognition, neurocomputation and neurotechnology. We interface between neuroscience, biological sciences, computer science, engineering and the AI and data science community at the University of Cambridge. We enable collaborations across Cambridge’s cross-disciplinary research community in discovery neuroscience that have strong translational potential in the development of AI systems, neurotechnology solutions and clinical applications.