Neuropredictome

We currently have a working prototype of Neuropredictome, which is frequently being updated to improve usability and add additional features.

The latest version can be found under HERE.

Neuropredictome is a machine-learning based classification framework, optimized for the reliability and robustness of its associations. Search a phenotype, and Neuropredictome provides a comprehensive overview drawn from the world’s largest quality-curated datasets, results from which you can then compare to Neurosynth obtained meta-analyses.

• To avoid confirmation and reporting biases inherent in hypothesis-based research, initial linkages are pulled from UK-Biobank (N=19,831), which includes resting and task functional MRI as well as structural T1-weighted and diffusion tensor imaging, as well as 5,034 phenotypes.

• Because brain-based disorders often have important medical consequences for the rest of the body, and non-brain-based disorders often have important cognitive, psychiatric, and behavioral consequences. Thus, one unique feature of Neuropredictome is that interactions are considered without restricting assumptions to specific systems or field heuristics.

• Results generated by data-driven classifiers are then cross-validated, using deep-learning textual analyses, against 14,371 peer-reviewed research articles.

• Finally, structural equation modeling is used to identify driving influences between disparate signs and symptoms of a disease or phenotype.

For more information, see:

Sultan SF, Mujica-Parodi LR, Skiena S. Neuropredictome: A data-driven predictome linking neuroimaging to phenotype.