publications

2026

1. Learnable Protein Representations in Computational Biology for Predicting Drug-Target Affinity

   Rachit Kumar, Joseph Romano, and Marylyn Ritchie

   Journal of Cheminformatics, Jan 2026

   Abs DOI

   In this review, we discuss the various different types of learnable protein representations that have been used in computational biology, with a particular focus on representations that have been used in the paradigm of predicting drug-target affinity. We explore this from multiple perspectives: the source of protein information used, the training paradigms used in generating and applying such representations, and the types of (deep-learning-based) encoding or embedding methods that have been used to generate and operate on such representations. We focus on drug-target affinity due to its particular relevance and utility in the field of drug development and assessment, and we make suggestions for how drug-target affinity prediction methods development can be further improved by examining the current literature from the aforementioned perspectives. This survey thus serves as a valuable resource for researchers seeking to develop methods for predicting drug-target affinity by exploring how protein information has been used and could be used in effective ways to improve such predictions.

2. SAIGE-GPU — Accelerating Genome- and Phenome-Wide Association Studies Using GPUs

   Alex Rodriguez, Youngdae Kim, Tarak Nath Nandi, Karl Keat, Rachit Kumar, Mitchell Conery, Rohan Bhukar, Molei Liu, John Hessington, Ketan Maheshwari, and 14 more authors

   Bioinformatics, Jan 2026

   Abs DOI

   Genome-wide association studies (GWAS) at biobank scale are computationally intensive, especially for admixed populations requiring robust statistical models. SAIGE is a widely used method for generalized linear mixed-model GWAS but is limited by its CPU-based implementation, making phenome-wide association studies impractical for many research groups.We developed SAIGE-GPU, a GPU-accelerated version of SAIGE that replaces CPU-intensive matrix operations with GPU-optimized kernels. The core innovation is distributing genetic relationship matrix calculations across GPUs and communication layers. Applied to 2,068 phenotypes from 635,969 participants in the Million Veteran Program (MVP), including diverse and admixed populations, SAIGE-GPU achieved a 5-fold speedup in mixed model fitting on supercomputing infrastructure and cloud platforms. We further optimized the variant association testing step through multi-core and multi-trait parallelization. Deployed on Google Cloud Platform and Azure, the method provided substantial cost and time savings.Source code and binaries are available for download at https://github.com/saigegit/SAIGE/tree/SAIGE-GPU-1.3.3. A code snapshot is archived at Zenodo for reproducibility (DOI: [10.5281/zenodo.17642591]). SAIGE-GPU is available in a containerized format for use across HPC and cloud environments and is implemented in R/C\,++ and runs on Linux systems.Supplementary data are available at Bioinformatics online.

2025

1. The Moment AI Arrived in the Clinic: Insights from the SAIL 2025 Year in Review

   Pierre Elias, Kameron Collin Black, Payal Chandak, Elizabeth Healey, Rachit Kumar, Michelle M. Li, Hugo Morales, Brett Beaulieu Jones, and Emily Alsentzer

   NEJM AI, Oct 2025

   DOI
2. CASTER-DTA: Equivariant Graph Neural Networks for Predicting Drug–Target Affinity

   Rachit Kumar, Joseph D Romano, and Marylyn D Ritchie

   Briefings in Bioinformatics, Sep 2025

   Abs DOI

   Accurately determining the binding affinity of a ligand with a protein is important for drug design, development, and screening. With the advent of accessible protein structure prediction methods such as AlphaFold, predicted protein 3D structures are readily available; however, scalable methods for predicting binding affinity currently do not take full advantage of 3D protein information. Here, we present CASTER-DTA (Cross-Attention with Structural Target Equivariant Representations for Drug–Target Affinity), which uses an equivariant graph neural network (GNN) to learn more robust protein representations alongside a standard GNN to learn molecular representations to predict DTA. We augment these representations by incorporating an attention-based mechanism between protein residues and drug atoms to improve interpretability. We show that CASTER-DTA represents a state-of-the-art improvement on multiple benchmarks for predicting DTA, and that it generates novel insights for several related tasks. We then apply CASTER-DTA to create a large resource of the binding affinities of every drug approved by the U.S. Food and Drug Administration (FDA) against every protein in the human proteome and make these predictions freely available for download. We also make available a web server for researchers to apply a pretrained CASTER-DTA model for predicting binding affinities between arbitrary proteins and drugs.

3. Graph Neural Networks for Interpretable Biomedical Data Analysis in Genomics and Structural Biology

   Rachit Kumar

   University of Pennsylvania, Sep 2025

   Abs

   The quantity and scope of biomedical data has increased dramatically, and with such an increase has come a corresponding increase in the need for scalable methods for integrating such data. Simultaneously, ensuring the interpretability of such methods has similarly become critical to not only improving the prospect of translating such methods to a variety of domains, but also enabling their use in identifying new areas of inquiry and insight. We first provide an overview of the literature and past work of other researchers in two fields: (1) network-based multiomics analysis; and (2) learnable protein representations in structural biology with a focus on drug-target affinity. We find that graph neural networks play a prominent role in both of these paradigms but that many gaps remain in their effective utilization thereof, inspiring us to develop new methods that make use of graph neural networks more effectively in related paradigms. We then show how graph neural networks can be used in genomics data analysis by discussing our development of a graph neural network model for predicting disease from genomics data alone, showcasing this model in a case study of Alzheimer’s disease. To show that graph neural networks provide unique benefits, we highlight our model’s performance and inherent interpretability in identifying contributors to the development of Alzheimer’s disease. We subsequently discuss the development of a graph neural network that predicts drug-target affinity from protein and drug structures, represented as graphs. We further emphasize the value of using graph neural networks in this way by highlighting its utility and interpretability in applying it to several downstream tasks such as predicting protein mutation impact on drug binding and identifying binding residues of drugs as well as highlighting its ability to preferentially rank known drugs that target certain proteins. Our state-of-the-art results by using graph neural networks in these paradigms while maintaining high levels of interpretability demonstrate the potential of using graph neural networks across the biomedical data analysis spectrum, showcasing the inherent interpretability and power that can be gained when using graph-based methods for generating representations of complex biomedical data and analyzing such data.

4. Network-Based Analyses of Multiomics Data in Biomedicine

   Rachit Kumar, Joseph D. Romano, and Marylyn D. Ritchie

   BioData Mining, May 2025

   Abs DOI

   Network representations of data are designed to encode relationships between concepts as sets of edges between nodes. Human biology is inherently complex and is represented by data that often exists in a hierarchical nature. One canonical example is the relationship that exists within and between various -omics datasets, including genomics, transcriptomics, and proteomics, among others. Encoding such data in a network-based or graph-based representation allows the explicit incorporation of such relationships into various biomedical big data tasks, including (but not limited to) disease subtyping, interaction prediction, biomarker identification, and patient classification. This review will present various existing approaches in using network representations and analysis of data in multiomics in the framework of deep learning and machine learning approaches, subdivided into supervised and unsupervised approaches, to identify benefits and drawbacks of various approaches as well as the possible next steps for the field.

5. Quantitative Electroencephalogram and Machine Learning to Predict Expired Sevoflurane Concentration in Infants

   Rachit Kumar, Justin Skowno, Britta S. von Ungern-Sternberg, Andrew Davidson, Ting Xu, Jianmin Zhang, XingRong Song, Mazhong Zhang, Ping Zhao, Huacheng Liu, and 32 more authors

   Journal of Clinical Monitoring and Computing, Oct 2025

   Abs DOI

   Processed electroencephalography (EEG) indices used to guide anesthetic dosing in adults are not validated in young infants. Raw EEG can be processed mathematically, yielding quantitative EEG parameters (qEEG). We hypothesized that machine learning combined with qEEG can accurately classify expired sevoflurane concentrations in young infants. Knowledge from this may contribute to development of future infant-specific EEG algorithms. Frontal EEG collected from infants\,≤ 3 months were time-matched as one-minute epochs to expired sevoflurane (eSevo). Fifteen qEEG parameters were extracted from each epoch and eight machine learning models combined the qEEG to classify each epoch into one of four eSevo levels (%): 0.1–1.0, 1.0–2.1, 2.1–2.9, and\,> 2.9. 64 epochs formed the post hoc SHAP dataset to determine the qEEG that contributed most to the model. The remaining epochs were randomly split 50 times into 80/20 training/testing sets. Accuracy and F1-score determined model performance. 42 infants provided 4574 epochs. The top classifiers K-nearest neighbors, default multi-layer perceptron, and support vector machine achieved 67.5–68.7% accuracy. Burst suppression ratio and entropy β were the top contributors to the models. Post hoc analysis performed without burst suppression ratio yielded similar prediction performance. In young infants, machine learning applied to qEEG predicted eSevo levels with moderate success. Burst suppression ratio, the most important contributor, represented an efficient EEG feature that encapsulated underlying EEG changes seen on other qEEG features. These results provided insight into EEG parameter selection and optimal machine learning models used for future development of infant-specific EEG algorithms.

6. Integrative Multi-Omics Approaches Identify Molecular Pathways and Improve Alzheimer’s Disease Risk Prediction

   Rasika Venkatesh, Katie M. Cardone, Yuki Bradford, Anni K. Moore, Rachit Kumar, Jason H. Moore, Li Shen, Dokyoon Kim, and Marylyn D. Ritchie

   Alzheimer’s & Dementia, Oct 2025

   Abs DOI

   INTRODUCTION Alzheimer’s disease (AD) is a complex neurodegenerative disorder with heterogeneous genetic and molecular underpinnings. Polygenic scores (PGS) capture little of this complexity. METHODS We conducted genome-, transcriptome-, and proteome-wide association studies (G/T/PWAS) on 15,480 individuals from the Alzheimer’s Disease Sequencing Project R4 (ADSP) to identify AD-associated signals, followed by pathway enrichment analysis. Integrative risk models (IRMs) were developed using genetically regulated components of gene and protein expression and clinical covariates. Elastic-net logistic regression and random forest classifiers were evaluated using standard metrics and compared against baseline PGS. RESULTS Known and novel signals were identified via G/T/PWAS. Enrichment analyses highlighted cholesterol and immune signaling pathways. The best-performing IRM, random forest with transcriptomic and covariate features, achieved area under the receiver operating characteristic (AUROC) of 0.703 and area under the precision-recall curve (AUPRC) of 0.622, significantly outperforming PGS and baseline models. DISCUSSION Integrating univariate discovery approaches with multivariate modeling enhances AD risk prediction and offers novel insights into underlying biological processes. Highlights Identified novel contributions to Alzheimer’s disease (AD) from a multi-omics perspective. Integrated genome-wide association studies (GWAS), transcriptome-wide association studies (TWAS), and proteome-wide association studies (PWAS) in a unified association study framework. Developed a method for predicting heritable risk of late-onset AD. Demonstrated that ancestry-aware modeling improves AD risk prediction accuracy.

2024

1. PGxQA: A Resource for Evaluating LLM Performance for Pharmacogenomic QA Tasks

   Karl Keat, Rasika Venkatesh, Yidi Huang, Rachit Kumar, Sony Tuteja, Katrin Sangkuhl, Binglan Li, Li Gong, Michelle Whirl-Carrillo, Teri E. Klein, and 2 more authors

   In Biocomputing 2025, Oct 2024

   DOI
2. A Comprehensive Bibliometric Analysis: Celebrating the Thirtieth Anniversary of the Pacific Symposium on Biocomputing

   Rachit Kumar, Rasika Venkatesh, David Y. Zhang, Teri E. Klein, and Marylyn D. Ritchie

   In Biocomputing 2025, Oct 2024

   DOI
3. Genetic Algorithm Selection of Interacting Features (GASIF) for Selecting Biological Gene-Gene Interactions

   Rachit Kumar, David Zhang, and Marylyn DeRiggi Ritchie

   In Proceedings of the Genetic and Evolutionary Computation Conference, Jul 2024

   Abs DOI

   Feature interactions are particularly useful in modeling biological effects, such as gene-gene interactions, but are difficult to model due to the exponential increase in the feature space. We present GASIF, a Genetic Algorithm that selects features and their interactions for the purposes of solving a supervised classification problem, designed for the identification of gene-gene interactions. GASIF works by constructing individuals with a collection of chromosomes that represent a subset of features and their interactions. It then determines individual fitness as a combination of the number of unique features used and the cross-validation performance of a logistic regression classifier trained on that feature subset with an ElasticNet penalty. A variety of intuitive operations are used to select, mate, and mutate individuals from generation to generation to limit the search space of features and interactions. We evaluate this Genetic Algorithm on a real-world dataset of human brain transcriptomic data from neuropathologically normal postmortem samples and pathologically confirmed late-onset Alzheimer’s disease individuals and determine the face validity of the gene-gene interactions that it identifies. Across multiple iterations of GASIF, we consistently identified the same features and interactions as most informative, all of which relate to genes known to be implicated in Alzheimer’s disease.

4. SABER: Statistical Identification of Loci of Interest in GWAS Summary Statistics Using a Bayesian Gaussian Mixture Model

   Rachit Kumar, Rasika Venkatesh, and Marylyn D. Ritchie

   AMIA Joint Summits on Translational Science proceedings. AMIA Joint Summits on Translational Science, Jul 2024

   Abs

   Genome-wide association studies (GWAS) remain a popular method for identifying novel genetic associations with human phenotypes and have provided many insights into the etiology of many diseases. However, GWAS provide limited support for how a genetic association might contribute to disease due to inherent limitations, such as linkage disequilibrium. As such, many methods that operate on GWAS summary statistics have been developed to generate evidence for functional pathways or for variants of interest, but they require defining the genomic region bounds for loci of interest. At present, there are limited methods for determining these bounds in a rigorous, reproducible way. We present a novel statistical method, Statistical Analysis for Bayesian Estimation of Regions (SABER), that uses Bayesian Gaussian mixture models to reproducibly generate ratios that quantify whether particular genomic positions represent the bounds of loci of interest and can be used to delineate genomic regions for downstream analyses.

5. Known and Novel Genes Significant to AD across Tissues May Help to Uncover Comprehensive Changes of Disease

   Anni Moore, Yuki Bradford, Rachit Kumar, Rasika Venkatesh, Manu Shivakumar, Jingxuan Bao, Li Shen, Dokyoon Kim, and Marylyn D. Ritchie

   Alzheimer’s & Dementia, Jul 2024

   Abs DOI

   Background Alzheimer’s disease (AD) is neurodegenerative disease brought on by a combination of changes in multiple pathways that conglomerate to promote disease progression. AD often occurs alongside comorbid diseases, most often immune or vascular in nature, which have been shown to further increase AD risk. We previously showed that known AD variants also associate with secondary diseases in these categories, including rheumatoid arthritis, ischemic myocardial infarction, and both Type 1 and Type 2 diabetes. To better understand which pathways play a significant role in AD etiology and how comorbid diseases may contribute to risk through shared pathways, we aimed to identify gene expression changes associated with AD status in tissues throughout the body. Method We conducted a transcriptome-wide association study (TWAS) using genotypes from the Alzheimer’s Disease Sequencing Project (ADSP) including 11,074 cases and 14,310 controls from multiple ancestry groups and reference expression-trait quantitative loci (eQTLs) from 49 tissues in Genotype-Tissue Expression (GTEx) Project (v8). Using an ensemble approach we first predicted genetically regulated gene expression (GREx) and gene-associations per tissue within each of the 49 tissues available in GTEx using PrediXcan. Secondly, to better predict cross-tissue signals we used MulTiXcan with GREx to test for gene-associations across tissues. Result Both known and novel genes appeared in the single tissue and cross tissue gene associations. Previously AD-associated genes PET117, RAB35, NECTIN2, APOC1, and TOMM40 were significantly associated (p<1.51e-7) with AD across tissues, with many of these known to lie within the APOE region. Additionally, genes previously unreported with AD, but involved in relevant disease processes including macro-autophagy (SUPT20H), nucleosome binding (HMGN2), histone acetylation (KAT14), DNA damage repair (HFM1), and immune response (TSPAN2) showed significance. Many of these novel genes have previously been associated with immune, vascular and other neurological diseases, many of which are better known to appear as AD comorbidities. Conclusion In addition to known APOE region genes, novel genes with implications relevant to AD etiology and ties to other immune, vascular, and neurological diseases also appeared significant across tissues. Next we will investigate how these gene expression changes apply to pathways within relevant tissues.

6. Biomechanical Analysis of Complications Following T10-Pelvis Spinal Fusion: A Population Based Computational Study

   Austin Q. Nguyen, Christian Rodriguez, Rachit Kumar, Sachin Gupta, Dennis E. Anderson, and Comron Saifi

   Journal of Biomechanics, Mar 2024

   Abs DOI

   Proximal junctional kyphosis (PJK) and proximal junctional failure (PJF) are challenging complications of long fusion constructs for the treatment of adult spinal deformity. The objective of this study is to understand the biomechanical stresses proximal to the upper instrumentation of a T10-pelvis fusion in a large patient cohort. The pre-fusion models were subject-specific thoracolumbar spine models that incorporate the height, weight, spine curvature, and muscle morphology of 250 individuals from the Framingham Heart Study Multidetector CT Study. To create post-fusion models, the subject-specific models were further modified to eliminate motion between the intervertebral joints from T10 to the pelvis. OpenSim analysis tools were used to calculate the medial lateral shear force, anterior posterior shear force, and compressive force on the T9 vertebra during the static postures. Differences between pre-fusion and post-fusion T9 biomechanics were consistent between increased segmental mobility and unchanged segmental mobility conditions. For all static postures, compression decreased (p < 0. 0005). Anterior-posterior shear force significantly increased (p < 0. 0005) during axial twist and significantly increased (p < 0. 0005) during trunk flexion. Medial lateral shear force significantly increased (p < 0. 0005) during axial twist. This computational study provided the first use of subject-specific models to investigate the biomechanics of long spinal fusions. Patients undergoing T10-Pelvis fusion were predicted to have increased shear forces and decreased compressive force at the T9 vertebra, independent of change in segmental mobility. The computational model shows potential for the investigation of spinal fusion biomechanics to reduce the risk of PJK or PJF.

7. The Alzheimer’s Knowledge Base: A Knowledge Graph for Alzheimer Disease Research

   Joseph D. Romano, Van Truong, Rachit Kumar, Mythreye Venkatesan, Britney E. Graham, Yun Hao, Nick Matsumoto, Xi Li, Zhiping Wang, Marylyn D. Ritchie, and 2 more authors

   Journal of Medical Internet Research, Apr 2024

   Abs DOI

   BACKGROUND: As global populations age and become susceptible to neurodegenerative illnesses, new therapies for Alzheimer disease (AD) are urgently needed. Existing data resources for drug discovery and repurposing fail to capture relationships central to the disease’s etiology and response to drugs. OBJECTIVE: We designed the Alzheimer’s Knowledge Base (AlzKB) to alleviate this need by providing a comprehensive knowledge representation of AD etiology and candidate therapeutics. METHODS: We designed the AlzKB as a large, heterogeneous graph knowledge base assembled using 22 diverse external data sources describing biological and pharmaceutical entities at different levels of organization (eg, chemicals, genes, anatomy, and diseases). AlzKB uses a Web Ontology Language 2 ontology to enforce semantic consistency and allow for ontological inference. We provide a public version of AlzKB and allow users to run and modify local versions of the knowledge base. RESULTS: AlzKB is freely available on the web and currently contains 118,902 entities with 1,309,527 relationships between those entities. To demonstrate its value, we used graph data science and machine learning to (1) propose new therapeutic targets based on similarities of AD to Parkinson disease and (2) repurpose existing drugs that may treat AD. For each use case, AlzKB recovers known therapeutic associations while proposing biologically plausible new ones. CONCLUSIONS: AlzKB is a new, publicly available knowledge resource that enables researchers to discover complex translational associations for AD drug discovery. Through 2 use cases, we show that it is a valuable tool for proposing novel therapeutic hypotheses based on public biomedical knowledge.

8. Sociodemographic Factors and Research Experience Impact MD-PhD Program Acceptance

   Darnell K. Adrian Williams, Briana Christophers, Timothy Keyes, Rachit Kumar, Michael C. Granovetter, Alexandria Adigun, Justin Olivera, Jehron Pura-Bryant, Chynna Smith, Chiemeka Okafor, and 3 more authors

   JCI Insight, Feb 2024

   DOI

2023

1. Extending Tree-Based Automated Machine Learning to Biomedical Image and Text Data Using Custom Feature Extractors

   Rachit Kumar, Joseph Romano, Marylyn Ritchie, and Jason Moore

   In Proceedings of the Companion Conference on Genetic and Evolutionary Computation, Jul 2023

   Abs DOI

   Automated machine learning (AutoML) has allowed for many innovations in biomedical data science; however, most AutoML approaches do not support image or text data. To rectify this, we implemented four feature extractors in the Tree-based Pipeline Optimization Tool (TPOT) to make TPOT with Feature Extraction (TPOT-FE), an automated machine learning system that uses genetic programming (GP) to create ideal pipelines for a classification or regression task. These feature extractors enable TPOT-FE to build pipelines that can analyze non-tabular data, including text and images, which are increasingly common biomedical big data modalities that can contain rich quantities of information. We evaluate this approach on six image datasets and four text datasets, including three biomedical datasets, and show that TPOT-FE is able to consistently construct and optimize classification pipelines on all of the datasets.

2. Aging and Visual Presentations in MRI Environments

   Jenny Walker, Rachit Kumar, Marvin Hoo, and Mark E. Wheeler

   Proceedings of the International Symposium on Human Factors and Ergonomics in Health Care, Mar 2023

   Abs DOI

   As individuals age, their vision tends to decline. These changes are natural, but often neglected when designing tasks for both older and younger adults. For instance, many MRI machine set-ups include in-bore displays. Ongoing work in our lab suggests that some older adults have issues seeing stimuli on these displays if there is a visual noise component. This is a problem that did not occur with younger adult samples. Therefore, this work provides an example of how this concern was addressed using a psychophysical thresholding technique. We hope that our experience will inform others who are facing similar issues and/or seeking suggestions for improving their patient and participants’ scanning experiences.

2022

1. Intraoperative Hypotension Is Not Correlated With Acute Kidney Injury During Spinal Fusion Surgery

   Rachel Blue, Alexis Gutierrez, Hasan S. Ahmad, Maya Alexis, Rachit Kumar, Michael Spadola, Connor Wathen, Mitchell Weinstein, and Dmitriy Petrov

   International Journal of Spine Surgery, Dec 2022

   Abs DOI

   Background Intraoperative hypotension (IOH) has been found to be associated with organ damage, including cardiac injury and acute kidney injury (AKI). However, to our knowledge, this relationship has not been studied in a neurosurgery-specific patient population. In this report, we review our institutional experience to understand the magnitude of association between IOH in spinal fusion operations and incidence of postoperative AKI. Methods This retrospective cohort study included 910 patients who underwent posterior spinal fusion procedures performed in the prone position. Intraoperative variables collected and analyzed include minute-by-minute mean arterial pressure (MAP) from an arterial catheter, intermittent blood pressure cuff readings, volume of administered intravenous fluids, urine output, and all relevant vitals and administered medications. The electronic medical record was queried for additional patient data. IOH was defined as MAP <65 mm Hg for greater than 10 minutes. The primary endpoints of the study were presence and staging of AKI ( [Kidney Disease: Improving Global Outcomes] consensus classification), postoperative ileus, and postoperative troponin leak. Results Using a partial correlation analysis, no association was found between IOH metrics (IOH occurrence, IOH duration >10 minutes, and total IOH time) and any outcome metrics, including AKI, except for vasopressor usage and estimated blood loss. Patient age at surgery was not associated with any outcome variables. The lack of association between IOH and AKI contrasts with existing literature; this could be due to underlying differences in our patient population or could highlight a more complex relationship between IOH and AKI than previously understood. Conclusion Occurrence and duration of IOH were not associated with AKI, postoperative ileus, troponin leak, length of stay, or any other major outcome variables in spinal fusion patients. Clinical Relevance These findings depart from previous literature showing a correlation between IOH and AKI and provide level 3 evidence clinically relevant to spinal surgery. Further research is needed to better understand the exact nature of this relationship. Level of Evidence 3.

2. The Virtual Summer Research Program: Supporting Future Physician-Scientists from Underrepresented Backgrounds

   Briana Macedo, Briana Christophers, Rio Barrere-Cain, Yentli Soto Albrecht, Michael C. Granovetter, Rachit Kumar, Dania Daye, Elizabeth Bhoj, Lawrence Brass, and Jose Alexandre Rodrigues

   Journal of Clinical and Translational Science, Aug 2022

   Abs DOI

   Introduction: Physician-scientist training programs expect applicants to have had extensive research experience prior to applying. Even at the best of times, this leaves individuals from underserved and underrepresented backgrounds at a competitive disadvantage, especially those remote from major academic centers. The COVID-19 pandemic exacerbated that disadvantage by closing research laboratories and suspending summer research opportunities. Methods: The Virtual Summer Research Program (VSRP) was designed to combat this shortfall by helping participating students become better informed and better prepared for applying to MD/DO–PhD programs. 156 participants were recruited from historically black colleges and universities and from national organizations for underrepresented trainees. Participants were paired with medical school faculty members and current MD/DO–PhD students from 35 participating institutions. The program lasted for at least 4 weeks and included a short research project, interactive sessions, journal clubs, social events, and attendance at a regional American Physician Scientists Association conference. Results: In follow-up surveys, participants reported improvements in their science-related skills and in their confidence in becoming a physician-scientist, applying to training programs, and navigating mentorship relationships. A follow-up study completed one year later indicated that participants felt they had benefited from an enhanced skill set, long-term relationships with their mentors, and connections to the physician-scientist community at large. Discussion: The results suggest that VSRP met its primary goals, which were to provide a diverse group of trainees with mentors, provide skills and resources for MD/DO–PhD application and matriculation and to support the development of longitudinal relationships between VSRP mentees and APSA. VSRP provides an approach that can be applied at an even larger scale when the constraints caused by a global pandemic have lifted.

3. Environmental Enrichment Leads to Behavioral Circadian Shifts Enhancing Brain-Wide Functional Connectivity between Sensory Cortices and Eliciting Increased Hippocampal Spiking

   Francis A. M. Manno, Ziqi An, Rachit Kumar, Junfeng Su, Jiaming Liu, Ed X. Wu, Jufang He, Yanqiu Feng, and Condon Lau

   NeuroImage, May 2022

   Abs DOI

   Environmental enrichment induces widespread neuronal changes, but the initiation of the cascade is unknown. We ascertained the critical period of divergence between environmental enriched (EE) and standard environment (SE) mice using continuous infrared (IR) videography, functional magnetic resonance imaging (fMRI), and neuron level calcium imaging. Naïve adult male mice (n = 285, C57BL/6J, postnatal day 60) were divided into SE and EE groups. We assessed the linear time-series of motion activity using a novel structural break test which examined the dataset for change in circadian and day-by-day motion activity. fMRI was used to map brain-wide response using a functional connectome analysis pipeline. Awake calcium imaging was performed on the dorsal CA1 pyramidal layer. We found the preeminent behavioral feature in EE was a forward shift in the circadian rhythm, prolongation of activity in the dark photoperiod, and overall decreased motion activity. The crepuscular period of dusk was seen as the critical period of divergence between EE and SE mice. The functional processes at dusk in EE included increased functional connectivity in the visual cortex, motor cortex, retrosplenial granular cortex, and cingulate cortex using seed-based analysis. Network based statistics found a modulated functional connectome in EE concentrated in two hubs: the hippocampal formation and isocortical network. These hubs experienced a higher node degree and significant enhanced edge connectivity. Calcium imaging revealed increased spikes per second and maximum firing rate in the dorsal CA1 pyramidal layer, in addition to location (anterior-posterior and medial-lateral) effect size differences between EE and SE. The emergence of functional-neuronal changes due to enrichment consisted of enhanced hippocampal-isocortex functional connectivity and CA1 neuronal increased spiking linked to a circadian shift during the dusk period. Future studies should explore the molecular consequences of enrichment inducing shifts in the circadian period.

4. Structural and Functional Hippocampal Correlations in Environmental Enrichment During the Adolescent to Adulthood Transition in Mice

   Francis A. M. Manno, Rachit Kumar, Ziqi An, Muhammad Shehzad Khan, Junfeng Su, Jiaming Liu, Ed X. Wu, Jufang He, Yanqiu Feng, and Condon Lau

   Frontiers in Systems Neuroscience, Feb 2022

   Abs DOI

   Environmental enrichment is known to induce neuronal changes; however, the underlying structural and functional factors involved are not fully known and remain an active area of study. To investigate these factors, we assessed enriched environment (EE) and standard environment (SE) control mice over 30 days using structural and functional MRI methods. Naïve adult male mice (n = 30, ≈20 g, C57BL/B6J, postnatal day 60 initial scan) were divided into SE and EE groups and scanned before and after 30 days. Structural analyses included volumetry based on manual segmentation as well as diffusion tensor imaging (DTI). Functional analyses included seed-based analysis (SBA), independent component analysis (ICA), the amplitude of low-frequency fluctuation (ALFF), and fractional ALFF (fALFF). Structural results indicated that environmental enrichment led to an increase in the volumes of cornu ammonis 1 (CA1) and dentate gyrus. Structural results indicated changes in radial diffusivity and mean diffusivity in the visual cortex and secondary somatosensory cortex after EE. Furthermore, SBA and ICA indicated an increase in resting-state functional MRI (rsfMRI) functional connectivity in the hippocampus. Using parallel structural and functional analyses, we have demonstrated coexistent structural and functional changes in the hippocampal subdivision CA1. Future research should map alterations temporally during environmental enrichment to investigate the initiation of these structural and functional changes.

5. Quality Control Procedures for Genome-Wide Association Studies

   Van Q. Truong, Jakob A. Woerner, Tess A. Cherlin, Yuki Bradford, Anastasia M. Lucas, Chelsea C. Okeh, Manu K. Shivakumar, Daniel H. Hui, Rachit Kumar, Milton Pividori, and 5 more authors

   Current Protocols, Nov 2022

   Abs DOI

   Genome-wide association studies (GWAS) are being conducted at an unprecedented rate in population-based cohorts and have increased our understanding of the pathophysiology of many complex diseases. Regardless of the context, the practical utility of this information ultimately depends upon the quality of the data used for statistical analyses. Quality control (QC) procedures for GWAS are constantly evolving. Here, we enumerate some of the challenges in QC of genotyped GWAS data and describe the approaches involving genotype imputation of a sample dataset along with post-imputation quality assurance, thereby minimizing potential bias and error in GWAS results. We discuss common issues associated with QC of the GWAS data (genotyped and imputed), including data file formats, software packages for data manipulation and analysis, sex chromosome anomalies, sample identity, sample relatedness, population substructure, batch effects, and marker quality. We provide detailed guidelines along with a sample dataset to suggest current best practices and discuss areas of ongoing and future research. \copyright 2022 Wiley Periodicals LLC.

2021

1. Supporting Equity and Inclusion of Deaf and Hard-of-Hearing Individuals in Professional Organizations

   Julia Jones Huyck, Kelsey L. Anbuhl, Brad N. Buran, Henry J. Adler, Samuel R. Atcherson, Ozan Cakmak, Robert T. Dwyer, Morgan Eddolls, Fadhel El May, Juergen-Theodor Fraenzer, and 32 more authors

   Frontiers in Education, Oct 2021

   Abs DOI

   Disability is an important and often overlooked component of diversity. Individuals with disabilities bring a rare perspective to science, technology, engineering, mathematics, and medicine (STEMM) because of their unique experiences approaching complex issues related to health and disability, navigating the healthcare system, creatively solving problems unfamiliar to many individuals without disabilities, managing time and resources that are limited by physical or mental constraints, and advocating for themselves and others in the disabled community. Yet, individuals with disabilities are underrepresented in STEMM. Professional organizations can address this underrepresentation by recruiting individuals with disabilities for leadership opportunities, easing financial burdens, providing equal access, fostering peer-mentor groups, and establishing a culture of equity and inclusion spanning all facets of diversity. We are a group of deaf and hard-of-hearing (D/HH) engineers, scientists, and clinicians, most of whom are active in clinical practice and/or auditory research. We have worked within our professional societies to improve access and inclusion for D/HH individuals and others with disabilities. We describe how different models of disability inform our understanding of disability as a form of diversity. We address heterogeneity within disabled communities, including intersectionality between disability and other forms of diversity. We highlight how the Association for Research in Otolaryngology has supported our efforts to reduce ableism and promote access and inclusion for D/HH individuals. We also discuss future directions and challenges. The tools and approaches discussed here can be applied by other professional organizations to include individuals with all forms of diversity in STEMM.

2. Hearing Loss Impacts Gray and White Matter across the Lifespan: Systematic Review, Meta-Analysis and Meta-Regression

   Francis A.M. Manno, Raul Rodríguez-Cruces, Rachit Kumar, J. Tilak Ratnanather, and Condon Lau

   NeuroImage, May 2021

   Abs DOI

   Hearing loss is a heterogeneous disorder thought to affect brain reorganization across the lifespan. Here, structural alterations of the brain due to hearing loss are assessed by using unique effect size metrics based on Cohen’s d and Hedges’ g. These metrics are used to map coordinates of gray matter (GM) and white matter (WM) alterations from bilateral congenital and acquired hearing loss populations. A systematic review and meta-analysis revealed m = 72 studies with structural alterations measured with magnetic resonance imaging (MRI) (bilateral = 64, unilateral = 8). The bilateral studies categorized hearing loss into congenital and acquired cases (n = 7,445) and control cases (n = 2,924), containing 66,545 datapoint metrics. Hearing loss was found to affect GM and underlying WM in nearly every region of the brain. In congenital hearing loss, GM decreased most in the frontal lobe. Similarly, acquired hearing loss had a decrease in frontal lobe GM, albeit the insula was most decreased. In congenital, WM underlying the frontal lobe GM was most decreased. In congenital, the right hemisphere was more negatively impacted than the left hemisphere; however, in acquired, this was the opposite. The WM alterations most frequently underlined GM alterations in congenital hearing loss, while acquired hearing loss studies did not frequently assess the WM metric. Future studies should use the endophenotype of hearing loss as a prognostic template for discerning clinical outcomes.

3. Structural Alterations in a Rat Model of Short-Term Conductive Hearing Loss Are Associated With Reduced Resting State Functional Connectivity

   Francis A. M. Manno, Ziqi An, Rachit Kumar, Ed X. Wu, Jufang He, Yanqiu Feng, and Condon Lau

   Frontiers in Systems Neuroscience, Aug 2021

   Abs DOI

   Conductive hearing loss (CHL) results in attenuation of air conducted sound reaching the inner ear. How a change in air conducted sound alters the auditory system resulting in cortical alterations is not well understood. Here, we have assessed structural and functional magnetic resonance imaging (MRI) in an adult (P60) rat model of short-term conductive hearing loss (1 week). Diffusion tensor imaging (DTI) revealed fractional anisotropy (FA) and axial diffusivity alterations after hearing loss that circumscribed the auditory cortex (AC). Tractography found the lateral lemniscus tract leading to the bilateral inferior colliculus (IC) was reduced. For baseline comparison, DTI and tractography alterations were not found for the somatosensory cortex. To determine functional connectivity changes due to hearing loss, seed-based analysis (SBA) and independent component analysis (ICA) were performed. Short term conductive hearing loss altered functional connectivity in the AC and IC, but not the somatosensory cortex. The results present an exploratory neuroimaging assessment of structural alterations coupled to a change in functional connectivity after conductive hearing loss. The results and implications for humans consist of structural-functional brain alterations following short term hearing loss in adults.

2020

1. Challenges and Advice for MD/PhD Applicants Who Are Underrepresented in Medicine

   Carl Bannerman, Natalie Guzman, Rachit Kumar, Chelsea Nnebe, Jordan Setayesh, Amitej Venapally, and Jonathan H. Sussman

   Molecular Biology of the Cell, Nov 2020

   Abs DOI

   The importance of diversity is self-evident in medicine and medical research. Not only does diversity result in more impactful scientific work, but diverse teams of researchers and clinicians are necessary to address health disparities and improve the health of underserved communities. MD/PhD programs serve an important role in training physician-scientists, so it is critical to ensure that MD/PhD students represent diverse backgrounds and experiences. Groups who are underrepresented in medicine and the biomedical sciences include individuals from certain racial and ethnic backgrounds, individuals with disabilities, individuals from disadvantaged backgrounds, and women. However, underrepresented students are routinely discouraged from applying to MD/PhD programs due to a range of factors. These factors include the significant cost of applying, which can be prohibitive for many students, the paucity of diverse mentors who share common experiences, as well as applicants’ perceptions that there is inadequate support and inclusion from within MD/PhD programs. By providing advice to students who are underrepresented in medicine and describing steps programs can take to recruit and support minority applicants, we hope to encourage more students to consider the MD/PhD career path that will yield a more productive and equitable scientific and medical community.

2. Glymphatic Clearance of Simulated Silicon Dispersion in Mouse Brain Analyzed by Laser Induced Breakdown Spectroscopy

   Muhammad Shehzad Khan, Rachit Kumar, Sinai H. C. Manno, Irfan Ahmed, Alan Wing Lun Law, Raul R. Cruces, Victor Ma, William C. Cho, Shuk Han Cheng, and Condon Lau

   Heliyon, Apr 2020

   Abs DOI

   Silicon-based devices, such as neural probes, are increasingly used as electrodes for receiving electrical signals from neural tissue. Neural probes used chronically have been known to induce inflammation and elicit an immune response. The current study detects and evaluates silicon dispersion from a concentrated source in the mouse brain using laser induced breakdown spectroscopy. Element lines for Si (I) were found at the injection site at approximately 288 nm at 3hr post-implantation, even with tissue perfusion, indicating possible infusion into neural tissue. At 24hr and 1-week post-implantation, no silicon lines were found, indicating clearance. An isolated immune response was found by CD68 macrophage response at 24hr post injection. Future studies should measure chronic silicon exposure to determine if the inflammatory response is proportional to silicon administration. The present type of protocol, coupling laser induced breakdown spectroscopy, neuroimaging, histology, immunohistochemistry, and determination of clearance could be used to investigate the glymphatic system and different tissue states such as in disease (e.g. Alzheimer’s).