A precision medicine framework using artificial intelligence for the identification and confirmation of genomic biomarkers of response to an Alzheimer's disease therapy: Analysis of the blarcamesine (ANAVEX2-73) Phase 2a clinical study

Abstract

Introduction: The search for drugs to treat Alzheimer's disease (AD) has failed to yield effective therapies. Here we report the first genome-wide search for biomarkers associated with therapeutic response in AD. Blarcamesine (ANAVEX2-73), a selective sigma-1 receptor (SIGMAR1) agonist, was studied in a 57-week Phase 2a trial (NCT02244541). The study was extended for a further 208 weeks (NCT02756858) after meeting its primary safety endpoint.

Methods: Safety, clinical features, pharmacokinetic, and efficacy, measured by changes in the Mini-Mental State Examination (MMSE) and the Alzheimer's Disease Cooperative Study-Activities of Daily Living scale (ADCS-ADL), were recorded. Whole exome and transcriptome sequences were obtained for 21 patients. The relationship between all available patient data and efficacy outcome measures was analyzed with unsupervised formal concept analysis (FCA), integrated in the Knowledge Extraction and Management (KEM) environment.

Results: Biomarkers with a significant impact on clinical outcomes were identified at week 57: mean plasma concentration of blarcamesine (slope MMSE:P < .041), genomic variants SIGMAR1 p.Gln2Pro (ΔMMSE:P < .039; ΔADCS-ADL:P < .063) and COMT p.Leu146fs (ΔMMSE:P < .039; ΔADCS-ADL:P < .063), and baseline MMSE score (slope MMSE:P < .015). Their combined impact on drug response was confirmed at week 148 with linear mixed effect models.

Discussion: Confirmatory Phase 2b/3 clinical studies of these patient selection markers are ongoing. This FCA/KEM analysis is a template for the identification of patient selection markers in early therapeutic development for neurologic disorders.

Keywords: Alzheimer's disease; association rules; biomarker; genomic analysis; knowledge extraction management; machine learning; mixed effect models; precision medicine; unsupervised analysis.

Conflict of interest statement

Harald Hampel is an employee of Eisai Inc. and serves as Senior Associate Editor for the Journal Alzheimer's & Dementia; during the past three years he had received lecture fees from Servier, Biogen and Roche, research grants from Pfizer, Avid, and MSD Avenir (paid to the institution), travel funding from Eisai, Functional Neuromodulation, Axovant, Eli Lilly and company, Takeda and Zinfandel, GE‐Healthcare and Oryzon Genomics, consultancy fees from Qynapse, Jung Diagnostics, Cytox Ltd., Axovant, Anavex, Takeda and Zinfandel, GE Healthcare, Oryzon Genomics, and Functional Neuromodulation, and participated in scientific advisory boards of Functional Neuromodulation, Axovant, Eisai, Eli Lilly and company, Cytox Ltd., GE Healthcare, Takeda and Zinfandel, Oryzon Genomics and Roche Diagnostics. He is co‐inventor in the following patents as a scientific expert and has received no royalties: In Vitro Multiparameter Determination Method for The Diagnosis and Early Diagnosis of Neurodegenerative Disorders Patent Number: 8916388. In Vitro Procedure for Diagnosis and Early Diagnosis of Neurodegenerative Diseases Patent Number: 8298784. Neurodegenerative Markers for Psychiatric Conditions Publication Number: 20120196300. In Vitro Multiparameter Determination Method for The Diagnosis and Early Diagnosis of Neurodegenerative Disorders Publication Number: 20100062463. In Vitro Method for The Diagnosis and Early Diagnosis of Neurodegenerative Disorders Publication Number: 20100035286. In Vitro Procedure for Diagnosis and Early Diagnosis of Neurodegenerative Diseases Publication Number: 20090263822. In Vitro Method for The Diagnosis of Neurodegenerative Diseases Patent Number: 7547553. CSF Diagnostic in Vitro Method for Diagnosis of Dementias and Neuroinflammatory Diseases Publication Number: 20080206797. In Vitro Method for The Diagnosis of Neurodegenerative Diseases Publication Number: 20080199966. Neurodegenerative Markers for Psychiatric Conditions Publication Number: 20080131921. Coralie Williams, Adrien Etcheto, and Frédéric Parmentier are employed by Ariana Pharmaceuticals. Mohammad Afshar is employed by and a shareholder of Ariana Pharmaceuticals. Federico Goodsaid is employed by and a shareholder of Regulatory Pathfinders. Walter E. Kaufmann is employed by Anavex Life Sciences Corp. Christopher U. Missling is employed by and a shareholder of Anavex Life Sciences Corp.

© 2020 The Authors. Alzheimer's & Dementia: Translational Research & Clinical Interventions published by Wiley Periodicals, Inc. on behalf of Alzheimer's Association.

Figures

FIGURE 1

Overview of study design, data availability, and analytical methods. (a) Summary of consecutive clinical trials ANAVEX2‐73‐002 involving two parts: Part A over 5 weeks and Part B over 52 weeks, and the ANAVEX2‐73‐003 extension study over a planned 208‐week period immediately after the initial trial, with a cumulative 265 weeks. (b) Summary of integrated data sources. A total of 2527 features, from 1152 descriptors, were used for each subject, including 837 genomic sequences with amino acid changes, from a total of 27,155 annotated genes and 185 RNA expression profiles. Patient descriptors are shown in gray and outcomes in pink. (c) Classification of number of patient descriptors incorporated in the two analytical steps applied in the study: (1) Unsupervised FCA rule‐based analysis of response at week 57 and (2) longitudinal confirmation using mixed effect modeling of response over 148 weeks with markers found at week 57 to model two groups of ANAVEX2‐73 concentrations

FIGURE 2

Linear mixed effect (LME) models of change in Mini‐Mental State Examination (MMSE) and Alzheimer's Disease Co‐operative Study‐Activities of Daily Living scale (ADCS‐ADL) since baseline over 148 weeks. To analyze the effect of high concentration on outcome over time, significant fixed effect terms linked to concentration were kept in the model. This has the effect of “correcting” for all other parameters except concentration. Because part of the response signal is not explained by the model (random error), this “residual” was added to the adjusted response values. For each time point, the model generates an adjusted predicted outcome for each patient. This adjusted outcome includes the residual mentioned above. For each time point, a mean and standard deviation of the adjusted outcome were calculated for the 21 patients and represented as solid points and error bars linked by dotted lines. A, LME‐adjusted slopes for the high concentration (green) cohort versus low and medium concentration patient cohort (magenta) with time (in weeks) against adjusted change in MMSE. Average adjusted values with residuals at the population level were plotted at each time point (dotted line). B, LME‐adjusted slopes for the high concentration (green) cohort versus low and medium concentration patient cohort (magenta) with time (in weeks) against adjusted change in ADCS‐ADL. Average adjusted values with residuals at the population level were plotted at each time point (dotted line)

FIGURE 3

Unadjusted values of change in Alzheimer's Disease Co‐operative Study‐Activities of Daily Living scale (ADCS‐ADL) since baseline over 148 weeks. A, The plot presents (unadjusted) mean trajectories of change in ADCS‐ADL scores over interim 148 weeks of subgroups of ANAVEX2‐73‐002/003 patients depending on biomarker status (absent or present) and of patients given standard of care. The subgroups of ANAVEX2‐73‐002/003 patients are represented (blue, green, orange, pink, purple, and turquoise) with plot depending on identified biomarker and baseline criteria characteristics. B, Summary of the mean and standard deviations of different patient groups unadjusted change in ADCS‐ADL scores at 14 time points (weeks 0, 5, 17, 31, 41, 53, 57, 83, 96, 109, 122, 135, and 148). All patients in the ANAVEX2‐73 study are presented along with each patient subgroup depending on inclusion criteria. The standard of care mean changes from baseline for ADCS‐ADL scores were obtained from the literature with −6.7 point change in 1 year 42 and −10.5 change in 18 months/70 wk 43