• HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
Wednesday, November 18, 2026
BIOENGINEER.ORG
No Result
View All Result
  • Login
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
No Result
View All Result
Bioengineer.org
No Result
View All Result
Home NEWS Science News Technology

Likelihood Contrast Model Recovers Parkinson’s Trajectory Phenotypes From Sparse Data

by
November 18, 2026
in Technology
Reading Time: 3 mins read
0
Likelihood Contrast Model Recovers Parkinson's Trajectory Phenotypes From Sparse Data

Likelihood Contrast Model Recovers Parkinson's Trajectory Phenotypes From Sparse Data

Share on FacebookShare on TwitterShare on LinkedinShare on RedditShare on Telegram

A statistical modeling framework based on multiclass likelihood contrast classification can accurately identify mathematical trajectory phenotypes from sparse and irregular longitudinal clinical records, according to a study evaluating Parkinson’s disease motor assessment data. In clinical research, patient measurements are frequently collected at unaligned intervals and across irregular visits, presenting significant hurdles for standard classification systems that typically depend on data interpolation or extensive feature reduction. To address these operational challenges, researchers designed an interpretable multiclass likelihood contrast framework for classifying unaligned longitudinal trajectories directly rather than compressing individual patient histories into a restricted set of summary features.

The analytical framework, developed by researchers Sangam Pangeni, Md Rejuan Haque, and Fahad Mostafa from The University of New Mexico Gallup Campus, The Ohio State University Wexner Medical Center, and Arizona State University, relies on class-specific mixed-effects models. Under this formulation, the proposed approach fits class-specific models and classifies each held-out subject according to the largest marginal likelihood. The underlying statistical specification incorporates a quadratic mean trajectory across time while allowing for subject-specific random intercepts and random slopes to capture individual baseline differences and distinct rates of trajectory change.

To evaluate how effectively the method performs under challenging data constraints, the research team conducted both numerical simulation experiments and an empirical evaluation using Parkinson’s disease motor-UPDRS data. Within the Parkinson’s disease experiment, dense longitudinal histories were initially utilized to define three mathematical trajectory phenotypes, providing an objective baseline against which the recovery of trajectory patterns could be systematically tested.

To simulate severe real-world data missingness and test trajectory recovery, the investigators deliberately masked approximately ninety percent of the longitudinal observations from the patient histories. This substantial sparsification evaluated whether the multiclass likelihood contrast algorithm could reliably recover an individual’s trajectory phenotype when provided with only sparse, unaligned records.

Model performance was assessed using subject-level cross-validation and compared directly against an extensive suite of established machine learning and statistical classifiers. Specifically, the evaluation benchmarked the proposed framework against linear mixed-likelihood classification, functional k-nearest neighbors, random forest, support vector machine, polynomial-feature classification, and grid-based principal component analysis classifiers. By testing across these diverse alternative modeling approaches, the researchers determined how effectively the likelihood contrast formulation preserved temporal structure compared to conventional feature-based methods.

Despite the substantial ninety percent sparsification of the data, the proposed model achieved an accuracy of 0.8095, a macro-F1 score of 0.8160, a Matthews correlation coefficient of 0.7107, and a macro-area under the receiver operating characteristic curve of 0.9268. Furthermore, the model exhibited consistent classification performance across different data splits, recording a validation macro-F1 score of 0.7980 alongside a held-out test macro-F1 score of 0.8160, indicating stable generalization without overfitting to specific partitions.

These findings demonstrate that multiclass likelihood contrast learning can preserve informative temporal structure in sparse and irregular longitudinal records without reducing individual trajectories to a small number of summary features. However, the authors emphasized that the Parkinson’s disease findings represent the recovery of data-derived mathematical trajectory phenotypes rather than prospective clinical prediction or diagnostic validation. The research team concluded that further evaluation in larger independent cohorts with clinically defined outcomes and multimodal longitudinal biomarkers is needed to establish genuine clinical utility.

Subject of Research: Multiclass likelihood contrast classification of sparse longitudinal clinical trajectories

Article Title: Parkinson's disease phenotype recovery from sparse longitudinal trajectories using multiclass likelihood contrast classification

Article References: Pangeni, S., Haque, M. R., & Mostafa, F. (2026). Parkinson's disease phenotype recovery from sparse longitudinal trajectories using multiclass likelihood contrast classification. Frontiers in Artificial Intelligence, 9, Article 1968393. https://doi.org/10.3389/frai.2026.1968393

Image Credits: AI Generated

DOI: 10.3389/frai.2026.1968393

Keywords: multiclass likelihood contrast, longitudinal trajectories, Parkinson's disease, motor-UPDRS, sparse longitudinal records, trajectory phenotyping, mixed-effects models, quadratic mean trajectory, random intercepts and slopes, subject-level cross-validation, linear mixed-likelihood classification, functional k-nearest neighbors, random forest, support vector machine, polynomial-feature classification, principal component analysis, Matthews correlation coefficient

News Source: Diana Fleming. (November 18, 2026). Likelihood Contrast Model Recovers Parkinson’s Trajectory Phenotypes From Sparse Data. Scienmag.

Tags: functional k-nearest neighborslinear mixed-likelihood classificationlongitudinal trajectoriesMatthews correlation coefficientmixed-effects modelsmotor-UPDRSmulticlass likelihood contrastParkinson’s diseasepolynomial-feature classificationprincipal component analysisquadratic mean trajectoryRandom Forestrandom intercepts and slopessparse longitudinal recordssubject-level cross-validationsupport vector machinetrajectory phenotyping
Share12Tweet7Share2ShareShareShare1

Related Posts

Interpretable Model Identifies Acute Stroke Patients at Risk of Early Epileptiform Discharges

Interpretable Model Identifies Acute Stroke Patients at Risk of Early Epileptiform Discharges

November 18, 2026
Modified Halpern Proximal Algorithm Solves Multi-Structured Problems in Curved Hadamard Spaces

Modified Halpern Proximal Algorithm Solves Multi-Structured Problems in Curved Hadamard Spaces

November 18, 2026

Multi-Batch Quantum Scheduling Reduces Execution Costs on IBM Torino Processor

November 18, 2026

Kinematically Controlled 3D Printing Transforms Polymer Stringing into Suspended Tissue Scaffolds

November 18, 2026

POPULAR NEWS

  • Electroacupuncture Mobilizes Unique CD34-Positive Stem-Like Cells in Healthy Dogs

    Electroacupuncture Mobilizes Unique CD34-Positive Stem-Like Cells in Healthy Dogs

    29 shares
    Share 12 Tweet 7
  • Specialists Outline Cervical Cancer Screening Principles for Resource-Limited Settings

    29 shares
    Share 12 Tweet 7
  • Modified Halpern Proximal Algorithm Solves Multi-Structured Problems in Curved Hadamard Spaces

    29 shares
    Share 12 Tweet 7
  • Machine Learning Outperforms Pharmacokinetic Models Predicting Initial Vancomycin Concentrations in Septic Patients

    29 shares
    Share 12 Tweet 7

About

We bring you the latest biotechnology news from best research centers and universities around the world. Check our website.

Follow us

Recent News

Surgeons Perform Hybrid Thoracoscopic and Endoscopic Enucleation for Rare Esophageal Foregut Cyst

Vine Tea Transcriptome Analysis Identifies Key Regulators of High Dihydromyricetin Biosynthesis

Carbapenem-Resistant Klebsiella Infections Surge 72 Percent Across Europe Since 2019

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 88 other subscribers
  • Contact Us

Bioengineer.org © Copyright 2023 All Rights Reserved.

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • Homepages
    • Home Page 1
    • Home Page 2
  • News
  • National
  • Business
  • Health
  • Lifestyle
  • Science

Bioengineer.org © Copyright 2023 All Rights Reserved.