Causal feature selection using a knowledge graph combining structured knowledge from the biomedical literature and ontologies: A use case studying depression as a risk factor for Alzheimer's disease.

BACKGROUND: Causal feature selection is essential for estimating effects from observational data. Identifying confounders is a crucial step in this process. Traditionally, researchers employ content-matter expertise and literature review to identify confounders. Uncontrolled confounding from unidentified confounders threatens validity, conditioning on intermediate variables (mediators) weakens estimates, and conditioning on common effects (colliders) induces bias. Additionally, without special treatment, erroneous conditioning on variables combining roles introduces bias. However, the vast literature is growing exponentially, making it infeasible to assimilate this knowledge. To address these challenges, we introduce a novel knowledge graph (KG) application enabling causal feature selection by combining computable literature-derived knowledge with biomedical ontologies. We present a use case of our approach specifying a causal model for estimating the total causal effect of depression on the risk of developing Alzheimer's disease (AD) from observational data. METHODS: We extracted computable knowledge from a literature corpus using three machine reading systems and inferred missing knowledge using logical closure operations. Using a KG framework, we mapped the output to target terminologies and combined it with ontology-grounded resources. We translated epidemiological definitions of confounder, collider, and mediator into queries for searching the KG and summarized the roles played by the identified variables. We compared the results with output from a complementary method and published observational studies and examined a selection of confounding and combined role variables in-depth. RESULTS: Our search identified 128 confounders, including 58 phenotypes, 47 drugs, 35 genes, 23 collider, and 16 mediator phenotypes. However, only 31 of the 58 confounder phenotypes were found to behave exclusively as confounders, while the remaining 27 phenotypes played other roles. Obstructive sleep apnea emerged as a potential novel confounder for depression and AD. Anemia exemplified a variable playing combined roles. CONCLUSION: Our findings suggest combining machine reading and KG could augment human expertise for causal feature selection. However, the complexity of causal feature selection for depression with AD highlights the need for standardized field-specific databases of causal variables. Further work is needed to optimize KG search and transform the output for human consumption.

Complex Walking Tasks and Risk for Cognitive Decline in High Functioning Older Adults.

BACKGROUND: Performance on complex walking tasks may provide a screen for future cognitive decline. OBJECTIVE: To identify walking tasks that are most strongly associated with subsequent cognitive decline. METHODS: Community-dwelling older adults with Modified Mini-Mental State (3MS) >85 at baseline (n = 223; mean age = 78.7, 52.5% women, 25.6% black) completed usual-pace walking and three complex walking tasks (fast-pace, narrow-path, visuospatial dual-task). Slope of 3MS scores for up to 9 subsequent years (average = 5.2) were used to calculate a cognitive maintainer (slope ≥0) or decliner (slope <0) outcome variable. Logistic regression models assessed associations between gait speeds and being a cognitive decliner. A sensitivity analysis in a subsample of individuals (n = 66) confirmed results with adjudicated mild cognitive impairment (MCI) or dementia at 8-9 years post-walking assessment. RESULTS: Cognitive decliners were 52.5% of the sample and on average were slower for all walking tasks compared to maintainers. In models adjusted for demographic and health variables, faster fast-pace (OR = 0.87 per 0.1 m/s, 95% CI: 0.78, 0.97) and dual-task (OR = 0.84 per 0.1 m/s, 95% CI: 0.73, 0.96) gait speeds were associated with lower likelihood of being a cognitive decliner. Usual-pace gait speed was not associated (OR = 0.96 per 0.1 m/s, 95% CI: 0.85, 1.08). Results were nearly identical in analyses with adjudicated MCI or dementia as the outcome. CONCLUSION: Fast-pace and dual-task walking may provide simple and effective tools for assessing risk for cognitive decline in older individuals with high cognitive function. Such screening tools are important for strategies to prevent or delay onset of clinically meaningful change.