Neurodevelopmental heterogeneity and computational approaches for understanding autism

Research output: Contribution to journalReview article

Abstract

In recent years, the emerging field of computational psychiatry has impelled the use of machine learning models as a means to further understand the pathogenesis of multiple clinical disorders. In this paper, we discuss how autism spectrum disorder (ASD) was and continues to be diagnosed in the context of its complex neurodevelopmental heterogeneity. We review machine learning approaches to streamline ASD’s diagnostic methods, to discern similarities and differences from comorbid diagnoses, and to follow developmentally variable outcomes. Both supervised machine learning models for classification outcome and unsupervised approaches to identify new dimensions and subgroups are discussed. We provide an illustrative example of how computational analytic methods and a longitudinal design can improve our inferential ability to detect early dysfunctional behaviors that may or may not reach threshold levels for formal diagnoses. Specifically, an unsupervised machine learning approach of anomaly detection is used to illustrate how community samples may be utilized to investigate early autism risk, multidimensional features, and outcome variables. Because ASD symptoms and challenges are not static within individuals across development, computational approaches present a promising method to elucidate subgroups of etiological contributions to phenotype, alternative developmental courses, interactions with biomedical comorbidities, and to predict potential responses to therapeutic interventions.

Original languageEnglish (US)
Article number63
JournalTranslational psychiatry
Volume9
Issue number1
DOIs
StatePublished - Dec 1 2019

Fingerprint

Autistic Disorder
Aptitude
Psychiatry
Comorbidity
Phenotype
Autism Spectrum Disorder
Machine Learning
Therapeutics

PubMed: MeSH publication types

  • Journal Article
  • Research Support, N.I.H., Extramural
  • Research Support, Non-U.S. Gov't
  • Review

Cite this

Neurodevelopmental heterogeneity and computational approaches for understanding autism. / Jacob, Suma; Wolff, Jason J.; Steinbach, Michael S.; Doyle, Colleen B.; Kumar, Vipan; Elison, Jed T.

In: Translational psychiatry, Vol. 9, No. 1, 63, 01.12.2019.

Research output: Contribution to journalReview article

@article{4458c5c5da6549c8b7f41e62eed15ff9,
title = "Neurodevelopmental heterogeneity and computational approaches for understanding autism",
abstract = "In recent years, the emerging field of computational psychiatry has impelled the use of machine learning models as a means to further understand the pathogenesis of multiple clinical disorders. In this paper, we discuss how autism spectrum disorder (ASD) was and continues to be diagnosed in the context of its complex neurodevelopmental heterogeneity. We review machine learning approaches to streamline ASD’s diagnostic methods, to discern similarities and differences from comorbid diagnoses, and to follow developmentally variable outcomes. Both supervised machine learning models for classification outcome and unsupervised approaches to identify new dimensions and subgroups are discussed. We provide an illustrative example of how computational analytic methods and a longitudinal design can improve our inferential ability to detect early dysfunctional behaviors that may or may not reach threshold levels for formal diagnoses. Specifically, an unsupervised machine learning approach of anomaly detection is used to illustrate how community samples may be utilized to investigate early autism risk, multidimensional features, and outcome variables. Because ASD symptoms and challenges are not static within individuals across development, computational approaches present a promising method to elucidate subgroups of etiological contributions to phenotype, alternative developmental courses, interactions with biomedical comorbidities, and to predict potential responses to therapeutic interventions.",
author = "Suma Jacob and Wolff, {Jason J.} and Steinbach, {Michael S.} and Doyle, {Colleen B.} and Vipan Kumar and Elison, {Jed T.}",
year = "2019",
month = "12",
day = "1",
doi = "10.1038/s41398-019-0390-0",
language = "English (US)",
volume = "9",
journal = "Translational Psychiatry",
issn = "2158-3188",
publisher = "Nature Publishing Group",
number = "1",

}

TY - JOUR

T1 - Neurodevelopmental heterogeneity and computational approaches for understanding autism

AU - Jacob, Suma

AU - Wolff, Jason J.

AU - Steinbach, Michael S.

AU - Doyle, Colleen B.

AU - Kumar, Vipan

AU - Elison, Jed T.

PY - 2019/12/1

Y1 - 2019/12/1

N2 - In recent years, the emerging field of computational psychiatry has impelled the use of machine learning models as a means to further understand the pathogenesis of multiple clinical disorders. In this paper, we discuss how autism spectrum disorder (ASD) was and continues to be diagnosed in the context of its complex neurodevelopmental heterogeneity. We review machine learning approaches to streamline ASD’s diagnostic methods, to discern similarities and differences from comorbid diagnoses, and to follow developmentally variable outcomes. Both supervised machine learning models for classification outcome and unsupervised approaches to identify new dimensions and subgroups are discussed. We provide an illustrative example of how computational analytic methods and a longitudinal design can improve our inferential ability to detect early dysfunctional behaviors that may or may not reach threshold levels for formal diagnoses. Specifically, an unsupervised machine learning approach of anomaly detection is used to illustrate how community samples may be utilized to investigate early autism risk, multidimensional features, and outcome variables. Because ASD symptoms and challenges are not static within individuals across development, computational approaches present a promising method to elucidate subgroups of etiological contributions to phenotype, alternative developmental courses, interactions with biomedical comorbidities, and to predict potential responses to therapeutic interventions.

AB - In recent years, the emerging field of computational psychiatry has impelled the use of machine learning models as a means to further understand the pathogenesis of multiple clinical disorders. In this paper, we discuss how autism spectrum disorder (ASD) was and continues to be diagnosed in the context of its complex neurodevelopmental heterogeneity. We review machine learning approaches to streamline ASD’s diagnostic methods, to discern similarities and differences from comorbid diagnoses, and to follow developmentally variable outcomes. Both supervised machine learning models for classification outcome and unsupervised approaches to identify new dimensions and subgroups are discussed. We provide an illustrative example of how computational analytic methods and a longitudinal design can improve our inferential ability to detect early dysfunctional behaviors that may or may not reach threshold levels for formal diagnoses. Specifically, an unsupervised machine learning approach of anomaly detection is used to illustrate how community samples may be utilized to investigate early autism risk, multidimensional features, and outcome variables. Because ASD symptoms and challenges are not static within individuals across development, computational approaches present a promising method to elucidate subgroups of etiological contributions to phenotype, alternative developmental courses, interactions with biomedical comorbidities, and to predict potential responses to therapeutic interventions.

UR - http://www.scopus.com/inward/record.url?scp=85061026164&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85061026164&partnerID=8YFLogxK

U2 - 10.1038/s41398-019-0390-0

DO - 10.1038/s41398-019-0390-0

M3 - Review article

C2 - 30718453

AN - SCOPUS:85061026164

VL - 9

JO - Translational Psychiatry

JF - Translational Psychiatry

SN - 2158-3188

IS - 1

M1 - 63

ER -