About

Joseph T. Chang is the James A. Attwood Professor of Statistics and Data Science and the chairman of the Department of Statistics and Data Science at Yale University. His academic role includes teaching courses such as Probability and Bayesian Statistics. Professor Chang's research interests span several areas within Probability and Statistics, with a particular focus on stochastic processes, Bayesian inference, Monte Carlo simulation methods, evolutionary tree reconstruction, and statistical genetics. He actively collaborates with scientists on various projects, notably working with investigators from the Yale Child Study Center on autism research. His work integrates advanced statistical methodologies with practical applications in genetics and evolutionary studies.

Research topics

• Biology
• Audiology
• Neuroscience
• Developmental psychology
• Psychology
• Biochemistry
• Cell biology
• Medicine
• Cancer research

Selected publications

• L’esthétique percussive de la Famicom et de la Nintendo Entertainment System de 1983 à 1987

  Sciences du jeu · 2025-01-01

  articleOpen access1st authorCorresponding

  Cet article explore l'évolution de l'orchestration des percussions dans la musique composée pour la console Nintendo Entertainment System entre 1983 et 1987. Le canal de bruit de la NES est utilisé pour créer des effets sonores et des sons de percussion, et ses spécifications techniques sont bien documentées dans les publications pertinentes. Cependant, la recherche n’a pas encore exploré la manière dont chaque compositeur et compositrice ont manipulé ce canal pour simuler des instruments de percussion individuels. L'observation du timbre et de la durée du bruit révèle leurs choix en matière d'instruments de percussion en fonction de leur style. Dix bandes sonores de jeux vidéo sont étudiées à l'aide de Sonic Visualiser, un outil efficace pour étudier le timbre des sons. Des observations techniques et historiques permettent de découvrir les particularités stylistiques musicales de chaque compositeur et compositrice pour les instruments de percussion.

  PublisherOA PDFDOI

• Large Language Models for Accurate Mental Health Screening: Identifying Clinical Phenotypes in ED Healthcare Workers

  Research Square · 2025-07-18

  preprintOpen access
  PublisherOA PDFDOI

• Chemotherapy Coupled to Macrophage Inhibition Induces T-cell and B-cell Infiltration and Durable Regression in Triple-Negative Breast Cancer

  UNC Libraries · 2024-05-17

  articleOpen access

  Immunosuppressive elements within the tumor microenvironment, such as tumor-associated macrophages (TAM), can present a barrier to successful antitumor responses by cytolytic T cells. Here we employed preclinical syngeneic p53 null mouse models of triple-negative breast cancer (TNBC) to develop a treatment regimen that harnessed the immunostimulatory effects of low-dose cyclophosphamide coupled with the pharmacologic inhibition of TAMs using either a small-molecule CSF1R inhibitor or an anti-CSF1R antibody. This therapeutic combination was effective in treating several highly aggressive TNBC murine mammary tumor and lung metastasis models. Single-cell RNA sequencing characterized tumor-infiltrating lymphocytes including Th cells and antigen-presenting B cells that were highly enriched in responders to combination therapy. In one model that exhibited long-term posttreatment tumor regression, high-dimensional imaging techniques identified the close spatial localization of B220þ/CD86þ-activated B cells and CD4þ T cells in tertiary lymphoid structures that were present up to 6 weeks posttreatment. The transcriptional and metabolic heterogeneity of TAMs was also characterized in two closely related claudin-low/mesenchymal subtype tumor models with differential treatment responses. A murine TAM signature derived from the T12 model was highly conserved in human claudin-low breast cancers, and high expression of the TAM signature correlated with reduced overall survival in patients with breast cancer. This TAM signature may help identify human patients with claudin-low breast cancer that will benefit from the combination of cyclophosphamide and anti-CSF1R therapy. These studies illustrate the complexity of the tumor immune microenvironment and highlight different immune responses that result from rational immunotherapy combinations. Significance: Immunostimulatory chemotherapy combined with pharmacologic inhibition of TAMs results in durable treatment responses elicited by Th cells and B cells in claudin-low TNBC models.

  PublisherDOI

• Family history of psychiatric conditions and development of siblings of children with autism

  Autism Research · 2024-06-19 · 2 citations

  articleOpen access

  Younger siblings (SIBS) of children with autism exhibit a wide range of clinical and subclinical symptoms including social, cognitive, language, and adaptive functioning delays. Identifying factors linked with this phenotypic heterogeneity is essential for improving understanding of the underlying biology of the heterogenous outcomes and for early identification of the most vulnerable SIBS. Prevalence of neurodevelopmental (NDD) and neuropsychiatric disorders (NPD) is significantly elevated in families of children with autism. It remains unknown, however, if the family history associates with the developmental outcomes among the SIBS. We quantified history of the NDDs and NPDs commonly reported in families of children with autism using a parent interview and assessed autism symptoms, verbal, nonverbal, and adaptive skills in a sample of 229 SIBS. Multiple regression analyses were used to examine links between family history and phenotypic outcomes, whereas controlling for birth year, age, sex, demographics, and parental education. Results suggest that family history of schizophrenia, depression, anxiety, bipolar disorder, and intellectual disability associate robustly with dimensional measures of social affect, verbal and nonverbal IQ, and adaptive functioning in the SIBS. Considering family history of these disorders may improve efforts to predict long-term outcomes in younger siblings of children with autism and inform about familial factors contributing to high phenotypic heterogenetity in this cohort.

  PublisherOA PDFDOI

• Targeting prostate tumor low–molecular weight tyrosine phosphatase for oxidation-sensitizing therapy

  Science Advances · 2024-01-31 · 9 citations

  articleOpen access

  Protein tyrosine phosphatases (PTPs) play major roles in cancer and are emerging as therapeutic targets. Recent reports suggest low–molecular weight PTP (LMPTP)—encoded by the ACP1 gene—is overexpressed in prostate tumors. We found ACP1 up-regulated in human prostate tumors and ACP1 expression inversely correlated with overall survival. Using CRISPR-Cas9–generated LMPTP knockout C4-2B and MyC-CaP cells, we identified LMPTP as a critical promoter of prostate cancer (PCa) growth and bone metastasis. Through metabolomics, we found that LMPTP promotes PCa cell glutathione synthesis by dephosphorylating glutathione synthetase on inhibitory Tyr 270 . PCa cells lacking LMPTP showed reduced glutathione, enhanced activation of eukaryotic initiation factor 2–mediated stress response, and enhanced reactive oxygen species after exposure to taxane drugs. LMPTP inhibition slowed primary and bone metastatic prostate tumor growth in mice. These findings reveal a role for LMPTP as a critical promoter of PCa growth and metastasis and validate LMPTP inhibition as a therapeutic strategy for treating PCa through sensitization to oxidative stress.

  PublisherOA PDFDOI

• Digital biomarkers for diagnostic assessment of stress pathologies and neurocognitive performance

  Psychoneuroendocrinology · 2024-02-01

  article
  PublisherDOI

• Spectral top-down recovery of latent tree models

  Information and Inference A Journal of the IMA · 2023-04-27

  articleOpen access

  Abstract Modeling the distribution of high-dimensional data by a latent tree graphical model is a prevalent approach in multiple scientific domains. A common task is to infer the underlying tree structure, given only observations of its terminal nodes. Many algorithms for tree recovery are computationally intensive, which limits their applicability to trees of moderate size. For large trees, a common approach, termed divide-and-conquer, is to recover the tree structure in two steps. First, separately recover the structure of multiple, possibly random subsets of the terminal nodes. Second, merge the resulting subtrees to form a full tree. Here, we develop spectral top-down recovery (STDR), a deterministic divide-and-conquer approach to infer large latent tree models. Unlike previous methods, STDR partitions the terminal nodes in a non random way, based on the Fiedler vector of a suitable Laplacian matrix related to the observed nodes. We prove that under certain conditions, this partitioning is consistent with the tree structure. This, in turn, leads to a significantly simpler merging procedure of the small subtrees. We prove that STDR is statistically consistent and bound the number of samples required to accurately recover the tree with high probability. Using simulated data from several common tree models in phylogenetics, we demonstrate that STDR has a significant advantage in terms of runtime, with improved or similar accuracy.

  PublisherOA PDFDOI

• Developmental trajectories of the default mode, executive control, and salience networks from the third trimester through the newborn period

  bioRxiv (Cold Spring Harbor Laboratory) · 2022-09-28 · 5 citations

  preprintOpen access

  ABSTRACT Social cognition is critical to early learning. Functional imaging studies in adults and older children suggest the involvement of the default mode (DMN), executive control (ECN), and salience (SAL) networks in social cognition. These networks are vulnerable to environmental insults, and abnormalities of intra- and inter-network connectivity of the three are emerging as biomarkers of neurobehavioral disorders. However, the developmental trajectories of the DMN, ECN, and SAL across the third trimester of gestation and perinatal transition remain largely unknown. Employing resting-state functional MRI studies at 30-32, 34-36, and 40-44 weeks postmenstrual age (PMA), we tested the hypothesis that both intra- and inter-network functional connectivity in the DMN, ECN, and SAL develop across the 30-46 weeks PMA time interval in a longitudinal/cross-sectional sample of 84 fetuses and neonates. A secondary analysis addressed the impact of maternal mental health assessed at 28 weeks PMA on tri-network development from 30-46 weeks PMA. The DMN, ECN, and SAL develop across the third trimester of gestation and the first postnatal month. At the intra-network level, significant increases occurred between 36 to 44 weeks PMA for all three, with network strength values significantly different from 0 beginning at 40 weeks PMA for all. Functional connectivity increased less rapidly in the DMN than in the ECN and SAL networks, suggesting slower maturation of the network subserving social interactions. In contrast, significant inter-network DMN – ECN connectivity greater than 0 was found from 36 weeks PMA through the first postnatal month, suggesting the emergence of inter-network functional connectivity in the fetal brain. Finally, higher maternal mental health symptoms measured at the beginning of the third trimester negatively affected the developmental trajectory of the SAL network across the critical time interval of 30 weeks to 44 weeks PMA. Together, these data provide a framework to compare fetuses and neonates at risk for neurobehavioral disorders and assess the impact of the environment on the developing brain.

  PublisherOA PDFDOI

• Hypoconnectivity between anterior insula and amygdala associates with future vulnerabilities in social development in a neurodiverse sample of neonates

  Scientific Reports · 2022-09-28 · 15 citations

  articleOpen access

  Altered resting state functional connectivity (FC) involving the anterior insula (aINS), a key node in the salience network, has been reported consistently in autism. Here we examined, for the first time, FC between the aINS and the whole brain in a sample of full-term, postmenstrual age (PMA) matched neonates (mean 44.0 weeks, SD = 1.5) who due to family history have high likelihood (HL) for developing autism (n = 12) and in controls (n = 41) without family history of autism (low likelihood, LL). Behaviors associated with autism were evaluated between 12 and 18 months (M = 17.3 months, SD = 2.5) in a subsample (25/53) of participants using the First Year Inventory (FYI). Compared to LL controls, HL neonates showed hypoconnectivity between left aINS and left amygdala. Lower connectivity between the two nodes was associated with higher FYI risk scores in the social domain (r(25) = -0.561, p = .003) and this association remained robust when maternal mental health factors were considered. Considering that a subsample of LL participants (n = 14/41) underwent brain imaging during the fetal period at PMA 31 and 34 weeks, in an exploratory analysis, we evaluated prospectively development of the LaINS-Lamy connectivity and found that the two areas strongly coactivate throughout the third trimester of pregnancy. The study identifies left lateralized anterior insula-amygdala connectivity as a potential target of further investigation into neural circuitry that enhances likelihood of future onset of social behaviors associated with autism during neonatal and potentially prenatal periods.

  PublisherOA PDFDOI

• MYH9 Facilitates Cell Invasion and Radioresistance in Head and Neck Cancer via Modulation of Cellular ROS Levels by Activating the MAPK-Nrf2-GCLC Pathway

  Cells · 2022 · 41 citations

  • Cancer research
  • Biology
  • Cell biology

  The MYH9 (Myosin heavy chain 9), an architecture component of the actomyosin cytoskeleton, has been reported to be dysregulated in several types of cancers. However, how this molecule contributes to cancer development is still obscure. This study deciphered the molecular function of MYH9 in head and neck cancer (HNC). Cellular methods included clonogenic survival, wound-healing migration, and Matrigel invasion assays. Molecular techniques included RT-qPCR, western blot, luciferase reporter assays, and flow cytometry. Clinical association studies were undertaken by TCGA data mining, Spearman correlation, and Kaplan-Meier survival analysis. We found that MYH9 was overexpressed in tumors and associated with poor prognosis in HNC patients. MYH9 promoted cell motility along with the modulation of the extracellular matrix (fibronectin, ITGA6, fascin, vimentin, MMPs). Also, MYH9 contributed to radioresistance and was related to the expression of anti-apoptotic and DNA repairing molecules (XIAP, MCL1, BCL2L1, ATM, RAD50, and NBN). Mechanically, MYH9 suppressed cellular ROS levels, which were achieved by activating the pan-MAPK signaling molecules (Erk, p38, and JNK), the induction of Nrf2 transcriptional activity, and the up-regulation of antioxidant enzymes (GCLC, GCLM, GPX2). The antioxidant enzyme GCLC was further demonstrated to facilitate cell invasion and radioresistance in HNC cells. Thus, MYH9 exerts malignant functions in HNC by regulating cellular ROS levels via activating the MAPK-Nrf2-GCLC signaling pathway. As MYH9 contributes to radioresistance and metastasis, this molecule may serve as a prognostic biomarker for clinical application. Furthermore, an in vivo study is emergent to support the therapeutic potential of targeting MYH9 to better manage refractory cancers.

  PublisherOA PDFDOI

Recent grants

• NIH Grant P50HD093078

  NIH · $2.4M

Frequent coauthors

• Yuval Kluger

  Yale University

  24 shared

• Elena L. Grigorenko

  Yale University

  16 shared

• Ann‐Joy Cheng

  Linkou Chang Gung Memorial Hospital

  15 shared

• Sherman M. Weissman

  Yale University

  12 shared

• Peter E. Newburger

  University of Massachusetts Chan Medical School

  9 shared

• David Tuck

  VA Boston Healthcare System

  9 shared

• Yasuhiro Nakayama

  9 shared

• Abdelhakim Ben Nasr

  The University of Texas Medical Branch at Galveston

  9 shared