About

Yves A. Lussier, MD, FACMI, FAMIA, is the Chair of Biomedical Informatics at the University of Utah School of Medicine. He is a professional engineer and physician-scientist with international expertise in translational bioinformatics. Dr. Lussier is a pioneer in research informatics techniques, including systems biology, data representation through ontologies, and high-throughput methods in personalized medicine. His research group conducts hypothesis-driven computational modeling predictions in precision medicine, which are validated through in vitro, in vivo, and clinical trials. Prior to his current role, Dr. Lussier served as the Associate Vice President for Information Science and Chief Knowledge Officer of UArizona Health Sciences at the University of Arizona, where he developed programs in biomedical informatics, computational genomics, and precision health. He has also held faculty positions at the University of Illinois at Chicago, University of Chicago, and Columbia University. Throughout his career, he has contributed to advancing precision health approaches, big data analytical tools, and resource services, and has been recognized with numerous awards and honors, including induction as a Fellow of the American College of Medical Informatics and multiple outstanding publication awards.

Research topics

• Biology
• Medicine
• Immunology
• Computational biology
• Internal medicine
• Pathology

Selected publications

• Challenges encountered in executing family routines: a comparison between neurotypical children and those having attention deficit hyperactivity disorder or autism spectrum disorder

  Frontiers in Child and Adolescent Psychiatry · 2026-03-17

  articleOpen access

  Introduction: Daily routines play a central role in the child's development process and the establishment of harmonious family dynamics. However, many parents of children with attention-deficit/hyperactivity disorder (ADHD) and with autism spectrum disorder (ASD), report difficulties in establishing and maintaining routines. The aim of this study was to compare neurotypical, ASD, and ADHD children's performance on daily routines. Precisely, it aims to describe the difficulties, the impacts on the child and his family, and the nature of the difficulties. Method: = 1.78; 31.7% girls), including 104 children with ADHD, 49 children with ASD, and 52 neurotypical children. Analyses of covariance (ANCOVAs) were performed to compare the three groups of participants, controlling for children's age and gender, parental education and family structure. Results: The results show that families of children with ADHD generally perceive routines as more difficult than those of neurotypical children. Children with ADHD experience significantly more frustration than neurotypical children when performing routines. According to parents, these difficulties in carrying out daily routines adversely affect the family climate, making it more stressful and unpleasant. Discussion: In conclusion, understanding the differences in the difficulties faced by these three groups of children in carrying out their daily routines will make it easier to support families in implementing interventions that are better adapted to the child's specific needs.

  PublisherOA PDFDOI

• MPSE identifies newborns for whole genome sequencing within 48 h of NICU admission

  npj Genomic Medicine · 2025-06-12

  articleOpen access

  Identifying critically ill newborns who will benefit from whole genome sequencing (WGS) is difficult and time-consuming due to complex eligibility criteria and evolving clinical features. The Mendelian Phenotype Search Engine (MPSE) automates the prioritization of neonatal intensive care unit (NICU) patients for WGS. Using clinical data from 2885 NICU patients, we evaluated the utility of different machine learning (ML) classifiers, clinical natural language processing (CNLP) tools, and types of Electronic Health Record (EHR) data to identify sick newborns with genetic diseases. Our results show that MPSE can identify children most likely to benefit from WGS within the first 48 h after NICU admission, a critical window for maximally impactful care. Moreover, MPSE provided stable, robust means to identify these children using many combinations of classifiers, CNLP tools, and input data types-meaning MPSE can be used by diverse health systems despite differences in EHR contents and IT support.

  PublisherOA PDFDOI

• Paired-Sample and Pathway-Anchored MLOps Framework for Robust Transcriptomic Machine Learning in Small Cohorts: Model Classification Study

  JMIR Bioinformatics and Biotechnology · 2025-09-13

  articleOpen accessSenior author

  Background: Approximately 90% of the 65,000 human diseases are infrequent, collectively affecting ~400 million people, substantially limiting cohort accrual. This low prevalence constrains the development of robust transcriptome-based machine learning (ML) classifiers. Standard data-driven classifiers typically require cohorts of more than 100 participants per group to achieve clinical accuracy while managing high-dimensional input (~25,000 transcripts). These requirements are infeasible for microcohorts of ~20 individuals, where overfitting becomes pervasive. Objective: To overcome these constraints, we developed a classification method that integrates three enabling strategies: (i) paired-sample transcriptome dynamics, (ii) N-of-1 pathway-based analytics, and (iii) reproducible machine learning operations (MLOps) for continuous model refinement. Methods: Unlike ML approaches relying on a single transcriptome per subject, within-subject paired-sample designs-such as pre- versus post-treatment or diseased versus adjacent-normal tissue-effectively control intraindividual variability under isogenic conditions and within-subject environmental exposures (eg, smoking history, other medications, etc), improve signal-to-noise ratios, and, when pre-processed as single- studies (N-of-1), can achieve statistical power comparable with that obtained in animal models. Pathway-level N-of-1 analytics further reduces each sample's high-dimensional profile into ~4000 biologically interpretable features, annotated with effect sizes, dispersion, and significance. Complementary MLOp practices-automated versioning, continuous monitoring, and adaptive hyperparameter tuning-improve model reproducibility and generalization. Results: In two case studies of distinct diseases, human rhinovirus infection (HRV) versus matched healthy controls (n=16 training; n=3 test) and breast cancer tissues harboring TP53 or PIK3CA mutations versus adjacent normal tissue (n=27 training; n=9 test)-this approach achieved 90% precision and recall on an unseen breast cancer test set and 92% precision with 90% recall in rhinovirus fivefold cross-validation. Incorporating paired-sample dynamics boosted precision by up to 12% and recall by 13% in breast cancer and by 5% each in HRV. MLOps workflows yielded an additional ~14.5% accuracy improvement compared to traditional pipelines. Moreover, our method identified 42 critical gene sets (pathways) for rhinovirus response and 21 for breast cancer mutation status, selected as the most important features (mean decrease impurity) of the best-performing model, with retroactive ablation of top 20 features reducing accuracy by ~25%. Conclusions: These proof-of-concept results support the utility of integrating intrasubject dynamics, "biological knowledge"-based feature reduction (pathway-level feature reduction grounded in prior biological knowledge; eg, N-of-1-pathway analytics), and reproducible MLOp workflows can overcome cohort size limitations in infrequent disease, offering a scalable, interpretable solution for high-dimensional transcriptomic classification. Future work will extend these advances across various therapeutic and small cohort designs.

  PublisherDOI

• Développement d’une solution technologique pour soutenir l’employabilité de personnes autistes ou présentant une déficience intellectuelle

  Revue francophone de la déficience intellectuelle · 2025-01-01

  article

  L’intégration socioprofessionnelle représente un défi de taille pour les personnes autistes ou présentant une déficience intellectuelle (DI). Ces dernières années, les technologies ont démontré leur efficacité pour soutenir ces personnes dans leurs activités quotidiennes, notamment en milieu de travail. Cette étude, structurée en trois phases, vise à développer un assistant à la réalisation de tâches (ART) destiné à favoriser leur employabilité. En adoptant une approche centrée sur l’utilisateur, l’ART a été ajusté en fonction des rétroactions des participants, incluant des travailleurs autistes ou présentant une DI et des membres du personnel d’encadrement. Les résultats mettent en lumière les facteurs facilitants et contraignants liés à l’utilisation de l’ART dans un contexte d’emploi.

  PublisherDOI

• Enabling Transcriptome Classification in Micro-cohorts with Pathway-Anchoring and Single-Subject Studies

  Lecture notes in computer science · 2025-01-01 · 1 citations

  book-chapterSenior author
  PublisherDOI

• Soutenir les activités quotidiennes d’adultes présentant une déficience intellectuelle : démarche d’implantation technologique

  Revue francophone de la déficience intellectuelle · 2025-01-01

  article

  En milieu résidentiel, le recours aux technologies pour favoriser la réalisation d’activités quotidiennes (p. ex., cuisiner, faire la lessive) par les personnes présentant une déficience intellectuelle (DI) s’avère une solution prometteuse. Toutefois, pour que cette intervention procure les bénéfices escomptés, il s’avère essentiel d’ancrer son déploiement dans la réalité vécue par les participants. L’objectif de l’étude consiste à décrire, à l’aide d’un devis qualitatif longitudinal (trois temps de mesure), le processus d’implantation et d’utilisation des technologies par six adultes présentant une DI pour réaliser des activités quotidiennes. Les résultats issus des analyses thématiques mettent en lumière l’importance de personnaliser l’intervention technologique au profil de chacun et de favoriser leur participation lors de la programmation. L’étude révèle également que les personnes intervenantes occupent un rôle central dans le déploiement. Enfin, l’intervention technologique a produit plusieurs effets chez les participants, la majorité d’entre eux étant positifs (p. ex., amélioration d’habiletés en cuisine, en contexte de conversations et en gestion émotionnelle).

  PublisherDOI

• Évaluation de l’acceptabilité dans le cadre d’un déploiement technologique au sein d’un milieu résidentiel accueillant des adultes autistes avec/sans déficience intellectuelle

  Revue francophone de la déficience intellectuelle · 2025-01-01

  article1st authorCorresponding

  L’évolution rapide de l’ère numérique force les organisations à innover pour s’arrimer aux besoins émergents. Toutefois, l’intégration d’une nouvelle technologie peut entrainer des résistances auprès des parties prenantes impliquées. Conséquemment, il s’avère pertinent de mesurer leurs perceptions au fil du temps. L’objectif de cette étude consiste à examiner les perceptions des employés travaillant au sein d’un milieu résidentiel accueillant des adultes autistes avec/sans déficience intellectuelle au sujet de l’implantation de solutions technologiques (tablettes numériques et applications). Pour ce faire, un devis descriptif longitudinal à trois temps de mesure a été utilisé. Au total, 16 employés ont participé. Les résultats obtenus montrent que les perceptions des participants se sont modifiées positivement et négativement dans le temps, en fonction de leurs interactions avec les solutions technologiques implantées, permettant ainsi d’identifier des leviers d’action et des obstacles à la démarche de déploiement.

  PublisherDOI

• Immunogenetics associated with severe coccidioidomycosis

  UNC Libraries · 2025-07-31

  articleOpen access

  Disseminated coccidioidomycosis (DCM) is caused by Coccidioides, pathogenic fungi endemic to the southwestern United States and Mexico. Illness occurs in approximately 30% of those infected, less than 1% of whom develop disseminated disease. To address why some individuals allow dissemination, we enrolled patients with DCM and performed whole-exome sequencing. In an exploratory set of 67 patients with DCM, 2 had haploinsufficient STAT3 mutations, and defects in β-glucan sensing and response were seen in 34 of 67 cases. Damaging CLEC7A and PLCG2 variants were associated with impaired production of β-glucan-stimulated TNF-α from PBMCs compared with healthy controls. Using ancestry-matched controls, damaging CLEC7A and PLCG2 variants were overrepresented in DCM, including CLEC7A Y238* and PLCG2 R268W. A validation cohort of 111 patients with DCM confirmed the PLCG2 R268W, CLEC7A I223S, and CLEC7A Y238* variants. Stimulation with a DECTIN-1 agonist induced DUOX1/DUOXA1-derived hydrogen peroxide [H2O2] in transfected cells. Heterozygous DUOX1 or DUOXA1 variants that impaired H2O2 production were overrepresented in discovery and validation cohorts. Patients with DCM have impaired β-glucan sensing or response affecting TNF-α and H2O2 production. Impaired Coccidioides recognition and decreased cellular response are associated with disseminated coccidioidomycosis.

  PublisherDOI

• The AI Moonshot: What We Need and What We Do Not

  The Annals of Family Medicine · 2025-01-01 · 2 citations

  editorialOpen accessSenior author
  PublisherDOI

• Paired-Sample and Pathway-Anchored MLOps Framework for Robust Transcriptomic Machine Learning in Small Cohorts (Preprint)

  2025-07-17

  articleOpen access1st authorCorresponding

  <sec> <title>BACKGROUND</title> Ninety percent of the 65,000 human diseases are infrequent, collectively affecting ~ 400 million people, substantially limiting cohort accrual. This low prevalence constrains the development of robust transcriptome-based machine learning (ML) classifiers. Standard data-driven classifiers typically require cohorts of over 100 subjects per group to achieve clinical accuracy while managing high-dimensional input (~25,000 transcripts). These requirements are infeasible for micro-cohorts of ~20 individuals, where overfitting becomes pervasive </sec> <sec> <title>OBJECTIVE</title> To overcome these constraints, we developed a classification method that integrates three enabling strategies: (i) paired-sample transcriptome dynamics, (ii) N-of-1 pathway-based analytics, and (iii) reproducible machine learning operations (MLOps) for continuous model refinement. </sec> <sec> <title>METHODS</title> Unlike ML approaches relying on a single transcriptome per subject, within-subject paired-sample designs — such as pre- versus post-treatment or diseased versus adjacent-normal tissue —effectively control intra-individual variability under isogenic conditions and within-subject environmental exposures (e.g. smoking history, other medications, etc.), improve signal-to-noise ratios, and, when pre-processed as single-subject studies (N-of-1), can achieve statistical power comparable to that obtained in animal models. Pathway-level N-of-1 analytics further reduces each sample’s high-dimensional profile into ~4,000 biologically interpretable features, annotated with effect sizes, dispersion, and significance. Complementary MLOps practices—automated versioning, continuous monitoring, and adaptive hyperparameter tuning—improve model reproducibility and generalization. </sec> <sec> <title>RESULTS</title> In two case studies—human rhinovirus infection versus matched healthy controls (n=16 training; 3 test) and breast cancer tissues harboring TP53 or PIK3CA mutations versus adjacent normal tissue (n=27 training; 9 test)—this approach achieved 90% precision and recall on an unseen breast cancer test set and 92% precision with 90% recall in rhinovirus fivefold cross-validation. . Incorporating paired-sample dynamics boosted precision by up to 12% and recall by 13% in BC, and by 5% each in HRV. MLOps workflows yielded an additional ~14.5% accuracy improvement compared to traditional pipelines. Moreover, our method identified 42 critical gene-sets (pathways) for rhinovirus response and 21 for breast cancer mutation status, with retroactive ablation of top features reducing accuracy by ~25%. </sec> <sec> <title>CONCLUSIONS</title> These proof-of-concept results support the utility of integrating intra-subject dynamics, “biological knowledge”-based feature reduction (pathway-level feature reduction grounded in prior biological knowledge; e.g., N-of-1-pathways analytics), and reproducible MLOps workflows can overcome cohort-size limitations in infrequent disease, offering a scalable, interpretable solution for high-dimensional transcriptomic classification. Future work will extend these advances across various therapeutic and small-cohort designs. </sec> <sec> <title>CLINICALTRIAL</title> not applicable </sec>

  PublisherDOI

Recent grants

• Biophysical Imaging

  NIH · $36.3M · 2016–2027

• Systems-level genetic patterns underlying disseminated coccidioidomycosis in humans

  NIH · $244k · 2020–2021

• Immuno-Genetic Basis for Human Disseminated Coccidioidomycosis

  NIH · $2.3M · 2017–2022

• Systems-level genetic patterns underlying disseminated coccidioidomycosis in humans

  NIH · $204k · 2020–2022

Frequent coauthors

• Haiquan Li

  China University of Mining and Technology

  75 shared

• Colleen Kenost

  University of Utah

  71 shared

• Jianrong Li

  54 shared

• Joe G. N. Garcia

  University of Florida

  54 shared

• James L. Chen

  Tempus Labs (United States)

  48 shared

• Joanne Berghout

  Pfizer (United States)

  47 shared

• Andrew D. Boyd

  41 shared

• Vincent Gardeux

  École Polytechnique Fédérale de Lausanne

  38 shared

Education

• Post-Doctoral Fellowship, Canadian Research Council Fellow, Department of Biomedical Informatics

  Columbia University

  2001

• M.D.

  Université de Sherbrooke

  1989

• B. Engineering

  Université de Sherbrooke

  1985

Awards & honors

• Fellow of the American College of Medical Informatics (ACMI)
• 3 IBM Faculty Awards
• 1st recipient of the Columbia University Faculty Mentoring A…
• Ambassador for Health Sciences at the University of Sherbroo…
• 16 outstanding publication awards from the American Medical…