Ankita Saxena
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Ankita Saxena has a total of 23 co-authors. Here are the top co-authors by number of shared publications:
- Stephen V. Faraone
Department of Psychiatry and Behavioral Sciences, Norton College of Medicine, SUNY Upstate Medical University, Syracuse, USA
2 shared publications - Abdel Daoud
Department of Pathology, School of Medicine, Johns Hopkins University , Ross 659, 710 Rutand Avenue, Baltimore, MD 21205
1 shared publication - Ali Shervin Razavi
MD/PhD Student, Department of Neuroscience and Physiology, College of Graduate Studies and Norton College of Medicine, SUNY Upstate Medical University
1 shared publication - Catharina A. Hartman
Faculty of Medical Sciences, University of Groningen, The Netherlands
1 shared publication - Charlene E. Gamaldo
Johns Hopkins University, Baltimore, MD
1 shared publication Christopher Gaiteri
Department of Neuroscience and Physiology, State University of New York – Norton College of Medicine at Upstate Medical University, Syracuse, New York, USA
1 shared publication
Latest Publications
- Open Access
Multi-Stage Graph Attention Networks for Interpretable Alzheimer’s Disease Classification from Genome-Wide Association Data
Background: Genome-wide association studies have identified numerous variants associated with neuropsychiatric disorders. Although some significant loci can carry substantial risk, as in Alzheimer’s Disease, the remaining genetic variance is distributed across many small-effect loci. Polygenic risk scores (PRS) aggregate this risk but do not capture epistatic interactions, and offer limited biological interpretability and predictive accuracy. Computing gene level risk scores and integrating known or statistically validated gene-gene associations has the potential to increase interpretability and/or accuracy. Graph Neural Networks (GNNs) can leverage graph structured genetic data that models potential epistatic interactions to achieve these goals. Methods: We developed a three-stage Graph Attention Network (GAT) classifier using individual-level GWAS data from 7,358 participants across seven Alzheimer’s Disease Center cohorts. Nodes were defined as genes, with risk scores from AD and 11 genetically correlated phenotypes serving as features. We evaluated two graph construction strategies: gene co-expression networks derived from hippocampal transcriptomic data and curated pathway-based graphs. Additionally, a bilinear context module was incorporated to capture global gene-gene interactions beyond the graph topology. In Stage 1, a GNN encoder was trained on the graphs; Stage 2 injected PRS for non-coding SNPs after the encoder to better capture genetic risk via transfer learning, and Stage 3 applied adversarial training with gradient reversal for ancestry debiasing. GNN predictions were ensembled with whole-genome PRS using elastic net regression. Results: The best-performing GNN model — a GAT with bilinear context operating on the pathway graph — achieved an AUROC of 0.78 (95% CI: 0.75–0.80). Ensemble models combining Stage 2 or 3 GNN logits with whole-genome PRS achieved an AUROC of 0.82 (0.79–0.84), outperforming PRS alone (0.80). GxI attribution and additional explainability analyses revealed stage-specific biological signals, some of which re-capitulated known gene-phenotype associations and others which may reflect potential new areas of inquiry. Conclusion: A multi-stage GAT framework captures complementary, non-additive genetic signal that, when ensembled with PRS, improves the accuracy of AD classification. Post-hoc explainability analyses yield biologically interpretable gene networks, supporting the utility of graph-based deep learning for dissecting complex genetic architectures.
Open AccessMulti-Stage Graph Attention Networks for Interpretable Alzheimer’s Disease Classification from Genome-Wide Association Data
Background: Genome-wide association studies have identified numerous variants associated with neuropsychiatric disorders. Although some significant loci can carry substantial risk, as in Alzheimer’s Disease, the remaining genetic variance is distributed across many small-effect loci. Polygenic risk scores (PRS) aggregate this risk but do not capture epistatic interactions, and offer limited biological interpretability and predictive accuracy. Computing gene level risk scores and integrating known or statistically validated gene-gene associations has the potential to increase interpretability and/or accuracy. Graph Neural Networks (GNNs) can leverage graph structured genetic data that models potential epistatic interactions to achieve these goals. Methods: We developed a three-stage Graph Attention Network (GAT) classifier using individual-level GWAS data from 7,358 participants across seven Alzheimer’s Disease Center cohorts. Nodes were defined as genes, with risk scores from AD and 11 genetically correlated phenotypes serving as features. We evaluated two graph construction strategies: gene co-expression networks derived from hippocampal transcriptomic data and curated pathway-based graphs. Additionally, a bilinear context module was incorporated to capture global gene-gene interactions beyond the graph topology. In Stage 1, a GNN encoder was trained on the graphs; Stage 2 injected PRS for non-coding SNPs after the encoder to better capture genetic risk via transfer learning, and Stage 3 applied adversarial training with gradient reversal for ancestry debiasing. GNN predictions were ensembled with whole-genome PRS using elastic net regression. Results: The best-performing GNN model — a GAT with bilinear context operating on the pathway graph — achieved an AUROC of 0.78 (95% CI: 0.75–0.80). Ensemble models combining Stage 2 or 3 GNN logits with whole-genome PRS achieved an AUROC of 0.82 (0.79–0.84), outperforming PRS alone (0.80). GxI attribution and additional explainability analyses revealed stage-specific biological signals, some of which re-capitulated known gene-phenotype associations and others which may reflect potential new areas of inquiry. Conclusion: A multi-stage GAT framework captures complementary, non-additive genetic signal that, when ensembled with PRS, improves the accuracy of AD classification. Post-hoc explainability analyses yield biologically interpretable gene networks, supporting the utility of graph-based deep learning for dissecting complex genetic architectures.
- Open Access
Reward Dysfunction in General and Specific Psychopathology in Children and Adults
Background: Reward dysfunction has been implicated in many psychiatric disorders such as attention-deficit/hyperactivity disorders, depression, and substance use disorders. However, psychiatric comorbidities are common, and the specificity of reward dysfunction to individual psychopathologies is unknown. The objective of this study was to evaluate the association between reward functioning and specific or general psychopathologies. Methods: 1044 adults and their 1215 children (ages 6-12) completed various measures of the Positive Valence System domain from the Research Domain Criteria (RDoC), which included the Delayed and Probability Discounting Tasks, Energy Expenditure for Reward Task, and Iowa Gambling Task. Children also completed the Experienced Pleasure Scale for children, while adults completed the Temporal Experience of Pleasure Scale and the Behavioral Activation System tasks. Psychopathology was assessed using the Child Behavior Checklist (CBCL) for children and the Adult Self Report (ASR) for parents. Results: One general factor identified via principal factors factor analysis explained the majority of variance in psychopathology in both groups. Reward measures in both adults and children were significantly associated with general psychopathology as well as most specific psychopathologies. Some associations between reward and psychopathology did not hold following removal of general psychopathology; nonetheless, certain reward constructs were uniquely associated with specific disorder problems but not general psychopathology. Conclusion: Disorder problem specific associations with reward functioning can be identified after removal of comorbidity. General propensity toward psychopathology is significantly but not uniquely correlated with reward dysfunction. Altogether, this may have broader implications for future study of the role of reward in disease pathogenesis.
Open AccessReward Dysfunction in General and Specific Psychopathology in Children and Adults
Background: Reward dysfunction has been implicated in many psychiatric disorders such as attention-deficit/hyperactivity disorders, depression, and substance use disorders. However, psychiatric comorbidities are common, and the specificity of reward dysfunction to individual psychopathologies is unknown. The objective of this study was to evaluate the association between reward functioning and specific or general psychopathologies. Methods: 1044 adults and their 1215 children (ages 6-12) completed various measures of the Positive Valence System domain from the Research Domain Criteria (RDoC), which included the Delayed and Probability Discounting Tasks, Energy Expenditure for Reward Task, and Iowa Gambling Task. Children also completed the Experienced Pleasure Scale for children, while adults completed the Temporal Experience of Pleasure Scale and the Behavioral Activation System tasks. Psychopathology was assessed using the Child Behavior Checklist (CBCL) for children and the Adult Self Report (ASR) for parents. Results: One general factor identified via principal factors factor analysis explained the majority of variance in psychopathology in both groups. Reward measures in both adults and children were significantly associated with general psychopathology as well as most specific psychopathologies. Some associations between reward and psychopathology did not hold following removal of general psychopathology; nonetheless, certain reward constructs were uniquely associated with specific disorder problems but not general psychopathology. Conclusion: Disorder problem specific associations with reward functioning can be identified after removal of comorbidity. General propensity toward psychopathology is significantly but not uniquely correlated with reward dysfunction. Altogether, this may have broader implications for future study of the role of reward in disease pathogenesis.
- 5 April 2016
- Published by Wolters Kluwer Health in Neurology
- Vol. 86 (16_supplem)
Difficult Conversations in Neurology: Defining Medical Student Exposure and Educational Opportunities (P2.369)
- 5 April 2016
- Published by Wolters Kluwer Health in Neurology
- Vol. 86 (16_supplem)
Difficult Conversations in Neurology: Defining Medical Student Exposure and Educational Opportunities (P2.369)