National Institutes of Health
National Institute of Allergy and Infectious Diseases

Author Of 5 Presentations

Machine Learning/Network Science Late Breaking Abstracts

LB1184 - Molecular models of multiple sclerosis severity identify heterogeneity of pathogenic mechanisms (ID 1974)

Speakers
Presentation Number
LB1184
Presentation Topic
Machine Learning/Network Science

Abstract

Background

New drug development and clinical management of patients with chronic, polygenic diseases, such as Multiple Sclerosis (MS), are suboptimal due to our inability to measure putative pathogenic processes contributing to destruction of the central nervous system (CNS). While blood is an excellent source of biomarkers used in clinical practice for management of e.g., cardiovascular disease, cerebrospinal fluid (CSF) represents a biological fluid that bring us as close to the CNS tissue in living patients as possible. Therefore, CNS-specific biomarkers in CSF could provide an insight into pathogenic mechanisms underlying disease expression in patients, its temporal distribution, intra-individual heterogeneity, and ultimately lead to precision medicine-based polypharmacy regimens.

Objectives

We sought to determine if CSF biomarkers can be aggregated to predict future rates of MS progression and provide molecular insight into mechanisms of CNS destruction and repair.

Methods

Using DNA-based SOMAscan technology we blindly measured 1,305 CSF biomarkers in longitudinal CSF samples of untreated MS patients divided into training (N=129) and validation (N=64) cohorts. We used machine learning algorithms in the training cohort to generate models of MS severity while the independent validation cohort samples were used to assess the predictive power of the models.

Results

CSF biomarker-based random forest models, validated in an independent longitudinal cohort, were able to predict reliably future rates of disability progression in MS patients. We were able to rule out the hypothesis that neuro-degenerative aspects of MS represent “accelerated aging” and instead defined, on a molecular level, mechanisms that correlate with how fast MS patients lose central nervous system (CNS) tissue, reflected by volumetric brain imaging and by clinical disability outcomes.

Conclusions

Cluster analysis of identified biomarkers revealed intra-individual molecular heterogeneity of disease mechanisms that include both CNS- and immune-related pathways and may represent novel targets for drug development and personalized treatments that would inhibit MS progression.

Acknowledgments: The research was supported by the Intramural Research Program of the National Institutes of Health (NIH), National Institute of Allergy and Infectious Diseases (NIAID).

Collapse
Machine Learning/Network Science Poster Presentation

P0012 - Machine-learning optimized Combinatorial MRI scale (COMRISv2) correlates highly with MS disability (ID 1438)

Speakers
Presentation Number
P0012
Presentation Topic
Machine Learning/Network Science

Abstract

Background

Volumetric biomarkers derived from brain MRI correlate only mildly to moderately with disability scales in multiple sclerosis (MS) patients. We previously addressed this issue by employing machine learning (ML) to select semi-quantitative MRI (semi-qMRI) features and their weights in the Combinatorial MRI Scale (COMRISv1). COMRISv1 correlated strongly with physical and cognitive disability in an independent validation cohort.

Objectives

Building on this work, we aimed to test the hypothesis that more powerful ML algorithm (i.e., Random Forests; RF) to both fully quantitative (qMRI) and semi-qMRI biomarkers in COMRISv2 will outperform COMRISv1 in the ability to predict physical and cognitive disability in an independent cohort of MS patients.

Methods

The prospectively acquired MS patients (n=283, divided 2:1 into training and validation cohorts) underwent brain MRI imaging within days of clinical evaluation. Neurological examination was transcribed to NeurEx app that automatically computes disability scales. Semi-qMRI features were determined weekly by consensus of MS-trained neurologists, while qMRI features were computed by lesion-TOADS algorithm implemented to QMENTA platform. All measurements were acquired as part of an IRB-approved clinical protocol, and support was provided by the National Institute of Allergy and Infectious Disease Division of Intramural Research.

Results

All RF-based COMRISv2 models validated (p<0.0001 for all) in the independent cohort. The predictions were stronger for models of physical disability, from which the model based on granular CombiWISE scale achieved the highest correlation (Spearman Rho = 0.855; Linh’s concordance coefficient that reflects 1:1 concordance between predicted and measured outcome; CCC = 0.824). COMRISv2 model of cognitive disability predicted measured symbol digit modalities tests (SDMT) with Rho = 0.493. Unexpectedly, formal comparison of the models consisting only from qMRI or semi-qMRI features demonstrated stronger predictive power of the latter.

Conclusions

COMRISv2 predicts clinical outcomes with strong accuracy but models of physical disability favor qMRI biomarkers reflecting disease burden in the infratentorial compartment, which is currently not measurable as qMRI biomarkers with sufficient accuracy. Addition of qMRI biomarkers of telencephalon damage only strengthened the performance of cognitive disability COMRISv2 model.

Collapse
Biomarkers and Bioinformatics Poster Presentation

P0117 - Mystery of MRZ reaction in multiple sclerosis   (ID 1449)

Speakers
Presentation Number
P0117
Presentation Topic
Biomarkers and Bioinformatics

Abstract

Background

Background: Multiple Sclerosis (MS) is an immune mediated disease of the central nervous system (CNS). Epstein-Barr virus (EBV) is among the most well-established environmental risk factors in MS. MS patients also make low affinity antibodies against neurotrophic viruses such as measles (M), rubella (R), and zoster (Z), and they make these intrathecally. This phenomenon is quite specific for MS and it is called MRZ reaction. The pathophysiological role of MRZ reaction currently remains unclear.

Objectives

Objectives: Because EBV induces differentiation of infected B cells in antigen-nonspecific manner, the aim of this project is to test the hypothesis that the extent of MRZ reaction correlates with the amount of EBV infected B cells in the intrathecal compartment of MS patients and with the MS severity.

Methods

Methods: In a pilot cohort of 80 patients with broad range of MS severity, we blindly analyzed MRZ reaction by enzyme-linked immunosorbent assay (ELISA) applied to matched cerebrospinal fluid (CSF) and serum samples. To derive a single measure of the extent of MRZ reaction, we summed antibody indexes (AI) for each reaction component (M, R and Z) to MRZ score.

Results

Results: Upon unblinding the MS severity outcomes and diagnostic subgroups, the MRZ scores did not significantly correlate with B cell/plasma cell biomarkers known to be increased in MS CSF (i.e., B-cell maturation antigen/BCMA and IgG index) and were not significantly different between relapsing remitting, primary progressive and secondary progressive MS. Finally, the MRZ scores did not correlate with clinical outcomes of MS severity (MS disease severity scale: MS-DSS and brain damage severity: principal component of brain parenchymal fraction and the symbol digit modalities test normalized for patient’s age).

Conclusions

Conclusions: The study proved null hypothesis (i.e., MRZ reaction is not associated with intrathecal biomarkers of B cell/plasma cell expansion and with MS severity). The association between MRZ reaction and MS remains a mystery.

Acknowledgments: The research was supported by the Intramural Research Program of the NIH, NIAID.

Notice: This study involved patients and it was approved by NIAID IRB.

Collapse
Internet and Social Media Poster Presentation

P0664 - Smartphone based Symbol Digit Modalities Test reliably measures cognitive function in multiple sclerosis patients (ID 1822)

Speakers
Presentation Number
P0664
Presentation Topic
Internet and Social Media

Abstract

Background

Neurological examination is a powerful tool for diagnosing and measuring progression of neurodegenerative diseases. However, examinations are resource intensive and thus not practical for comprehensive measurement of neurological disability in chronic diseases. A remote digital solution may be more practical and particularly relevant now due to the ongoing COVID-19 pandemic.

Objectives

To clinically validate a digital adaptation of the Symbol Digit Modalities Test (SDMT) developed as part of a smartphone test suite replicating aspects of a neurological exam.

Methods

Participants consisted of healthy volunteers (HV; n=39) and multiple sclerosis (MS; n=154) patients, with a longitudinal subcohort that performed tests at home (n=15). During clinic visits, the smartphone test suite was administered alongside a full neurological exam. The smartphone SDMT featured randomization of symbol-digit code and testing sequence. The subjects also underwent written SDMT and brain MRI.

Results

Performance differed significantly between HV and MS cohorts (p<.0001). Performance on smartphone and written SDMT had strong evidence of association (R2=0.71, concordance coefficient CCC=0.69, p<.0001). Smartphone SDMT had higher criterion validity than written SDMT measured by correlation with T2 lesion load and brain atrophy. Correlations with NeurEx subdomains identified neurological functions involved in performance of each of the 3 functional cognitive tests. Correcting for these contributing non-cognitive disabilities generated linear regression models strongly predicting the amount of MS-related brain injury measured by volumetric MRI (R2 = 0.75, p < 0.0001 vs R2 = 0.62, p < 0.0001). Of the longitudinal cohort, 87% of patients demonstrated practice effects measured by non-linear regression. Averaging multiple sequential post-learning results significantly decreased threshold for identifying true test deteriorations on a patient level.

Conclusions

Smartphone SDMT allows for less resource intensive remote administration. The clinometric properties of smartphone SDMT compare to or outperform written SDMT. This study expands the validation of multiple neurological tests administered via smartphone and bring us closer to a patient-autonomous neurological examination.

Acknowledgments: The research was supported by the Intramural Research Program of the National Institutes of Health, National Institute of Allergy and Infectious Diseases (NIAID).

Notice: This study involved patients and was approved by NIAID IRB.

Collapse
Pathogenesis – Role of Glia Poster Presentation

P0958 - Drug library screen identifies inhibitors of toxic astrogliosis (ID 1006)

Speakers
Presentation Number
P0958
Presentation Topic
Pathogenesis – Role of Glia

Abstract

Background

Multiple sclerosis (MS) is a chronic neuroinflammatory disorder, in which activated immune cells directly or indirectly induce demyelination and axonal degradation. Inflammatory stimuli also change phenotype of astrocytes, making them neurotoxic. Resulting “toxic astrocyte” phenotype has been seen in animal models of neuroinflammation and in MS lesions. Proteins secreted by toxic astrocytes are elevated in the cerebrospinal fluid (CSF) of MS patients and reproducibly correlate with the rates of accumulation of neurological disability and brain atrophy. This suggests pathogenic role for toxic astrocytes in MS.

Objectives

Therefore, the goal of this study is to identify signaling pathways underlying induction of toxic astrogliosis and to detect therapeutic inhibitors for these processes.

Methods

Here we applied commercial library of small molecules (Selleck Chemicals LLC; 1431 drugs) that are either Food and Drug Administration (FDA) approved or in clinical development to an in vitro model of toxic astrogliosis to identify drugs and signaling pathways that inhibit inflammatory transformation of astrocytes to neuro-toxic phenotype.

Results

Inhibitors of three pathways related to the endoplasmic reticulum (ER) stress: 1. Proteasome, 2. Heat shock protein 90 (HSP90)- and 3. Mammalian target of rapamycin (mTOR) reproducibly decreased inflammation-induced conversion of astrocytes to toxic phenotype. Dantrolene, an anti-spasticity drug that inhibits calcium release through ryanodine receptors (RyR) expressed in the ER of CNS cells, also exerted inhibitory effect at in vivo-achievable concentrations. We also established CSF SERPINA3 as a relevant pharmacodynamic marker for inhibiting toxic astrocytes in clinical trials.

Conclusions

In conclusion, drug library screening provides mechanistic insight into generation of toxic astrocytes and identifies candidates for immediate proof-of-principle clinical trial(s).

Collapse

Presenter Of 1 Presentation

Machine Learning/Network Science Late Breaking Abstracts

LB1184 - Molecular models of multiple sclerosis severity identify heterogeneity of pathogenic mechanisms (ID 1974)

Speakers
Presentation Number
LB1184
Presentation Topic
Machine Learning/Network Science

Abstract

Background

New drug development and clinical management of patients with chronic, polygenic diseases, such as Multiple Sclerosis (MS), are suboptimal due to our inability to measure putative pathogenic processes contributing to destruction of the central nervous system (CNS). While blood is an excellent source of biomarkers used in clinical practice for management of e.g., cardiovascular disease, cerebrospinal fluid (CSF) represents a biological fluid that bring us as close to the CNS tissue in living patients as possible. Therefore, CNS-specific biomarkers in CSF could provide an insight into pathogenic mechanisms underlying disease expression in patients, its temporal distribution, intra-individual heterogeneity, and ultimately lead to precision medicine-based polypharmacy regimens.

Objectives

We sought to determine if CSF biomarkers can be aggregated to predict future rates of MS progression and provide molecular insight into mechanisms of CNS destruction and repair.

Methods

Using DNA-based SOMAscan technology we blindly measured 1,305 CSF biomarkers in longitudinal CSF samples of untreated MS patients divided into training (N=129) and validation (N=64) cohorts. We used machine learning algorithms in the training cohort to generate models of MS severity while the independent validation cohort samples were used to assess the predictive power of the models.

Results

CSF biomarker-based random forest models, validated in an independent longitudinal cohort, were able to predict reliably future rates of disability progression in MS patients. We were able to rule out the hypothesis that neuro-degenerative aspects of MS represent “accelerated aging” and instead defined, on a molecular level, mechanisms that correlate with how fast MS patients lose central nervous system (CNS) tissue, reflected by volumetric brain imaging and by clinical disability outcomes.

Conclusions

Cluster analysis of identified biomarkers revealed intra-individual molecular heterogeneity of disease mechanisms that include both CNS- and immune-related pathways and may represent novel targets for drug development and personalized treatments that would inhibit MS progression.

Acknowledgments: The research was supported by the Intramural Research Program of the National Institutes of Health (NIH), National Institute of Allergy and Infectious Diseases (NIAID).

Collapse