Uncovering precision phenotype-biomarker associations in traumatic brain injury using topological data analysis

Abstract

Background: Traumatic brain injury (TBI) is a complex disorder that is traditionally stratified based on clinical signs and symptoms. Recent imaging and molecular biomarker innovations provide unprecedented opportunities for improved TBI precision medicine, incorporating patho-anatomical and molecular mechanisms. Complete integration of these diverse data for TBI diagnosis and patient stratification remains an unmet challenge.

Methods and findings: The Transforming Research and Clinical Knowledge in Traumatic Brain Injury (TRACK-TBI) Pilot multicenter study enrolled 586 acute TBI patients and collected diverse common data elements (TBI-CDEs) across the study population, including imaging, genetics, and clinical outcomes. We then applied topology-based data-driven discovery to identify natural subgroups of patients, based on the TBI-CDEs collected. Our hypothesis was two-fold: 1) A machine learning tool known as topological data analysis (TDA) would reveal data-driven patterns in patient outcomes to identify candidate biomarkers of recovery, and 2) TDA-identified biomarkers would significantly predict patient outcome recovery after TBI using more traditional methods of univariate statistical tests. TDA algorithms organized and mapped the data of TBI patients in multidimensional space, identifying a subset of mild TBI patients with a specific multivariate phenotype associated with unfavorable outcome at 3 and 6 months after injury. Further analyses revealed that this patient subset had high rates of post-traumatic stress disorder (PTSD), and enrichment in several distinct genetic polymorphisms associated with cellular responses to stress and DNA damage (PARP1), and in striatal dopamine processing (ANKK1, COMT, DRD2).

Conclusions: TDA identified a unique diagnostic subgroup of patients with unfavorable outcome after mild TBI that were significantly predicted by the presence of specific genetic polymorphisms. Machine learning methods such as TDA may provide a robust method for patient stratification and treatment planning targeting identified biomarkers in future clinical trials in TBI patients.

Trial registration: ClinicalTrials.gov Identifier NCT01565551.

Conflict of interest statement

Competing Interests: Authors TCP, JP, PYL and GEC are current or former employees of Ayasdi, Inc. This commercial affiliation does not alter adherence to PLOS ONE policies on sharing data and materials. There are no other competing interests declared by the rest of the co-authors ARF, JLN, GTM, SRC, JKY, MDS, TI, ELY, PM, MJV, HFL, WAG, ABV, and DOO.

Figures

Fig 1

(A-E). Methodological work-flow for integrating diverse clinical TBI data. (A) Hypothetical example of a spatial bi-plot of individual patients (grey points) on 2 functional endpoints after TBI (GOS-E AND PTSD). The same approach can be applied to multiple metrics simultaneously using multivariate pattern detectors (e.g., principal component analysis) to produce a multivariate view of function. (B) In TRACK-TBI Pilot the same individuals (N = 586) were tracked prospectively across multiple domains (function, biomarkers, imaging) providing connections (lines) across domains to improve patient classification using the full syndromic space. (C) Multivariate pattern detection lens can be used to categorize (colors) patients across all domains. (D) Patient grouping by multivariate lens. (E) Topological visualization renders patient groups into individual nodes, colored by the multivariate lens. Edges (black lines) indicate individuals appearing in both groups producing a syndromic map of patient clusters.

Fig 2

(A-F). TBI CDE network topology identifies the PARP1 SNP as a candidate predictor of GOS-E deficits in mild TBI. Patients with TBI were mapped into a TDA network, highlighting color schemes for CT (A) and MRI (B) pathology and whether they had a confirmed diagnosis of PTSD (DSM IV) at 6 months post-TBI (C). Patients in the circled regions of the network were identified due to substantial dysfunction measured by the GOS-E both at 3 months (D) and 6 months (E) post-TBI, compared with other patients in the network with no CT pathology and no diagnosis of PTSD. Data-driven exploration of these patients in the network revealed a significant categorical enrichment for the PARP1 SNP (F), particularly the heterozygous allele (A/T). Heat map represents range of numerical values for each measure: Panels A-C yes (1 = red) vs, no (0 = blue); Panels D-E GOS-E range from less than 3 (blue) to 8 (red); Panel F PARP1 allele A/A = 1 = blue, A/T = 2 = yellow/green, T/T = 3 = red.

Fig 3

(A-B). Hypothesis testing of PARP1 genetic polymorphism influence on GOS-E deficits in mild TBI. GOS-E scores between 3 and 6 months post-TBI were plotted for patients who were CT negative (A) or CT positive (B), based on the SNP allele expressed (A/A = blue, A/T = yellow/green, T/T = red). Hypothesis testing of the interaction between CT pathology and the SNP allele over time revealed a significant 3-way interaction; however, no significance was detected at each time point individually. Only change in GOS-E over time was significant in patients with a negative head CT.*p < .05.