Adam Ferguson, UCSF Brain and Spinal Injury Center, San Francisco, USA

CLIMBER - Combining Literature-sourced Informatics, Meta-science, Bioinformatics Evidence Research

Funded in: 2021, 2022, 2023


Back to overview

Problem: Translation of findings from experimental models into clinical therapies
Target: Alignment of preclinical published and unpublished data
Goal: Unique set of data science questions for informative analysis

For spinal cord injury research to have impact, there must be translation of findings from animal models into clinical therapies. However, this translation is fundamentally limited by a gap in the way that scientific evidence is handled in animal models versus in human research. In animal models, science is judged based on the novelty of findings and elegance of the methods. In contrast, clinical research focuses on stability of therapies in the messy reality of human spinal cord injury. This conflict between 'novelty' vs. 'stability' presents a fundamental problem for translating decision-support from preclinical science into clinical therapies.

How do we align data from small, niche animal studies to human clinical data?

Currently animal-to-human data alignment is done randomly in a haphazard manner. In some instances, clinicians read the animal literature and then design clinical trials according to what they find most promising. Other times, animal researchers advocate for clinical trials based on their desire to see their intellectual property succeed in the clinic. No matter how clinical studies begin, regulatory agencies judge whether ensuing evidence warrants clinical implementation using a ranking system known as the Classes of Evidence.

Meta-analysis of published literature is the highest grade of clinical evidence. They are statistical studies of multiple conclusive studies pooled together about clinical assessments of therapeutics. (It may also be useful for bench-to-bedside translation.)

The group hypothesizes that meta-analysis of published data, combined with unpublished raw subject-level data (dark data) and advanced data science, will improve precision and reproducibility of potential therapeutics in preclinical models. This will enable robust benchmarking of therapies to test for therapeutics that predict good recovery trajectories in acute and chronic SCI (clinically called 'neuroconversion').