Zhi Zhao

• Nymoen, Hanne Marte Gjertsen; Zhao, Zhi; Horndalsveen, Henrik; Bjaanæs, Maria Moksnes; Grønberg, Bjørn Henning & Halvorsen, Tarje Onsøien [Show all 20 contributors for this article] (2026). Gut microbiota alterations in patients with non-small-cell lung cancer undergoing chemoradiotherapy. Molecular Oncology. ISSN 1574-7891. doi: 10.1002/1878-0261.70326. Full text in Research Archive Show summary

  Growing evidence suggests that gut microbial features may predict—and potentially modulate—responses to cancer therapy, particularly immunotherapy. However, the impact of concurrent chemoradiotherapy (CRT) on the gut microbiota remains less well‐understood. This knowledge gap is especially relevant in locally advanced non‐small‐cell lung cancer (NSCLC), where CRT typically precedes immunotherapy. We therefore investigated whether CRT alters gut microbial composition and whether such changes are associated with survival outcomes. Fecal samples were collected at three time points: prior to CRT (baseline), at completion of CRT, and immediately before initiation of consolidation immunotherapy, from 62 patients with locally advanced NSCLC. Microbiota profiling was performed using a qPCR‐based panel (PMP™) targeting 108 prevalent microbial taxa. Within‐sample (alpha) and between‐sample (beta) bacterial diversity were assessed across time points and clinical subgroups defined by antibiotic exposure and survival outcomes. Alpha diversity remained stable from baseline to completion of CRT (all P  > 0.60). Patients who received broad‐spectrum antibiotics during CRT had significantly lower alpha diversity compared with those who did not receive antibiotics ( P  = 0.017), reflecting a transient between‐group difference. Beta diversity differed modestly but significantly at baseline between survival groups (progression‐free survival ≥ 18 vs < 18 months; PERMANOVA R 2  = 0.028, P  = 0.045). The gut microbiota appears largely stable during CRT for locally advanced NSCLC but is susceptible to disruption by antibiotic use. Baseline beta diversity differences between survival groups may signal a potential role for gut microbial composition as a future preliminary biomarker and warrant further investigation.

• Willoch-Allen, Katarina; Zhao, Zhi; Duran, Pilar Ayuda; Mikaeloff, Flora; Struyf, Nona & Österroos, Albin [Show all 13 contributors for this article] (2026). A multi-center study on the consistency of drug sensitivity testing in patients with acute myeloid leukemia. NPJ precision oncology. 10(1). doi: 10.1038/s41698-026-01480-4. Full text in Research Archive Show summary

  Clinical implementation of patient cell-based drug testing is hampered by inconsistencies and irreproducible results. Using linear mixed models and drug response data from four cohorts of patients with acute myeloid leukemia (n = 993 samples), we identified sources of variability in ex vivo drug sensitivity testing across the studies. We identified several experimental factors, including the time until sample was tested, the type of culture media and viability assay, which led to systematic differences in drug sensitivity profiles. We further investigated the effects of sample types (fresh or frozen) and disease status (diagnosis, relapse, or refractory), and observed that peripheral blood samples resulted in higher drug sensitivity levels, compared to those derived from bone marrow. Technical factors, such as the type of plate reader and centrifugation procedure, further contributed to the response variability depending on the chosen drug response metric. These factors need to be standardized for improved consistency in functional precision medicine studies.

• becerra, Rafael Romero; Zhao, Zhi; Nebdal, Daniel Johan Hammer; Müller, Elisabeth; Bergholtz, Helga & Norum, Jens Henrik [Show all 7 contributors for this article] (2025). Improved analysis of in vivo drug combination experiments with a comprehensive statistical framework and web-tool. Nature Communications. 16(1). doi: 10.1038/s41467-025-65218-9. Full text in Research Archive Show summary

  Drug combination therapy is often required to overcome the limited benefits of monotherapy in cancer treatment. While several tools exist for in vitro drug synergy screening and assessment, there is a lack of integrated methods for statistical analysis of in vivo combination experiments. To fill this gap, we present SynergyLMM, a comprehensive modeling and design framework for evaluating drug combination effects in preclinical in vivo studies. Unlike other methods, SynergyLMM accommodates complex experimental designs, including multi-drug combinations, and offers practical options for statistical analysis of both synergy and antagonism through longitudinal drug interaction analysis, including model diagnostics and statistical power analysis. These functionalities allow researchers to optimize study designs and determine an appropriate number of animals and follow-up time points required to achieve sufficient synergy and statistical power. SynergyLMM is implemented as an easy-to-use web-application, making it widely accessible for researchers without programming skills. We demonstrate the versatility and added value of SynergyLMM through its applications to various experimental setups and treatment experiments with chemo-, targeted- and immunotherapy. These case studies showcase its potential to improve robustness, statistical rigor and consistency of preclinical drug combination results, enabling a faster and safer transition from preclinical to clinical testing.

• Castel, Charlotte Jocelyne Yvonne; Zhao, Zhi & Thoresen, Magne (2025). Comparing LASSO and IPF-LASSO for multi-modal data: variable selection with Type I error control. Journal of Statistical Computation and Simulation. ISSN 0094-9655. 95(10), p. 2204–2218. doi: 10.1080/00949655.2025.2487503. Full text in Research Archive Show summary

  Variable selection in high-dimensional regression models is challenging. Thus, developing stable and reliable methods for variable selection is essential. Omics data, a common source of high-dimensional data, brings the added complexity of integrating diverse genomic layers into the analysis. The IPF-LASSO model has previously addressed this by applying distinct penalty parameters for each data modality. However, incorporating heterogeneous data layers into variable selection with Type I error control remains an open problem. To address this, we applied stability selection to control the number of false positives in both IPF-LASSO and standard LASSO models. Our study aimed to compare the two methods, investigating whether introducing different penalty parameters per data modality enhances statistical power while controlling false positives. Two high-dimensional data structures were investigated in simulations, one with independent data and the other with correlated data. We also applied the models to breast cancer treatment data, where IPF-LASSO identified relevant clinical variables.

• Zhao, Zhi; Zobolas, John; Zucknick, Manuela & Aittokallio, Tero Antero (2024). Tutorial on survival modeling with applications to omics data. Bioinformatics. ISSN 1367-4803. 40(3). doi: 10.1093/bioinformatics/btae132. Full text in Research Archive Show summary

  Motivation Identification of genomic, molecular and clinical markers prognostic of patient survival is important for developing personalized disease prevention, diagnostic and treatment approaches. Modern omics technologies have made it possible to investigate the prognostic impact of markers at multiple molecular levels, including genomics, epigenomics, transcriptomics, proteomics and metabolomics, and how these potential risk factors complement clinical characterization of patient outcomes for survival prognosis. However, the massive sizes of the omics datasets, along with their correlation structures, pose challenges for studying relationships between the molecular information and patients’ survival outcomes. Results We present a general workflow for survival analysis that is applicable to high-dimensional omics data as inputs when identifying survival-associated features and validating survival models. In particular, we focus on the commonly used Cox-type penalized regressions and hierarchical Bayesian models for feature selection in survival analysis, which are especially useful for high-dimensional data, but the framework is applicable more generally. Availability and implementation A step-by-step R tutorial using The Cancer Genome Atlas survival and omics data for the execution and evaluation of survival models has been made available at https://ocbe-uio.github.io/survomics.

• Farooqi, Saima Jamil; Zhao, Zhi; Øjlert, Åsa Kristina; Thunold, Solfrid; Juul, Hedvig Vidarsdotter & Bjaanæs, Maria Moksnes [Show all 10 contributors for this article] (2024). Serum cytokines as a biomarker for immune checkpoint inhibitor toxicity in patients with pleural mesothelioma. Frontiers in Immunology. 15. doi: 10.3389/fimmu.2024.1480183. Full text in Research Archive Show summary

  Background: Pleural mesothelioma (PM) is a rare cancer with a dismal prognosis. Dual immune checkpoint inhibitors have improved overall survival, but the rate of immune-related adverse events (irAEs) is high. Serum cytokines reflect systemic immune reactions and may serve as biomarkers for irAEs. Patients and methods: Patients with pleural mesothelioma treated with nivolumab and ipilimumab with or without UV1 vaccine in the NIPU study were included. Serum cytokine levels were measured by Bio-Plex Pro Human Cytokine Screening 48-Plex Panel Assay. Correlations between cytokine levels and irAEs were analyzed by generalized linear mixed models to identify potential diagnostic and predictive biomarkers. Results: Higher levels of MIG, eotaxin, MIP-1α, IP-10, TNF-α, MIP-1β, IL-4, MIF, IL-16, IL-2RA, SCGF.β and PDFG-BB at baseline are associated with increased risk of developing one or more irAEs. In particular, higher baseline levels of MIG are positively associated with thyroiditis and hypophysitis, and elevated levels of IP-10 and MIG to dermatitis. During the course of treatment, higher levels of MIG, eotaxin, MIF, TNF-α, MIP-1β, IL-4 and IL-16 are associated with an ongoing irAE. We found both predictive and diagnostic value of MIF with fatigue and of eotaxin with both colitis and pneumonitis. Higher levels of CTACK is associated with a lower risk of developing hepatitis, both before and after treatment. Conclusions: Elevated levels of certain cytokines, both before and after onset of treatment, correlate with specific irAEs in PM patients receiving ICIs. These cytokines may be used as biomarkers to predict and detect irAES.

• Ankill, Jørgen; Zhao, Zhi; Tekpli, Xavier; Kure, Elin Wenche Hegland; Kristensen, Vessela N. & Mathelier, Anthony [Show all 7 contributors for this article] (2024). Integrative pan-cancer analysis reveals a common architecture of dysregulated transcriptional networks characterized by loss of enhancer methylation. PLoS Computational Biology. ISSN 1553-734X. 20:e1012565(11), p. 1–33. doi: 10.1371/journal.pcbi.1012565. Full text in Research Archive Show summary

  Aberrant DNA methylation contributes to gene expression deregulation in cancer. However, these alterations’ precise regulatory role and clinical implications are still not fully understood. In this study, we performed expression-methylation Quantitative Trait Loci (emQTL) analysis to identify deregulated cancer-driving transcriptional networks linked to CpG demethylation pan-cancer. By analyzing 33 cancer types from The Cancer Genome Atlas, we identified and confirmed significant correlations between CpG methylation and gene expression (emQTL) in cis and trans, both across and within cancer types. Bipartite network analysis of the emQTL revealed groups of CpGs and genes related to important biological processes involved in carcinogenesis including proliferation, metabolism and hormone-signaling. These bipartite communities were characterized by loss of enhancer methylation in specific transcription factor binding regions (TFBRs) and the CpGs were topologically linked to upregulated genes through chromatin loops. Penalized Cox regression analysis showed a significant prognostic impact of the pan-cancer emQTL in many cancer types. Taken together, our integrative pan-cancer analysis reveals a common architecture where hallmark cancer-driving functions are affected by the loss of enhancer methylation and may be epigenetically regulated.

• Sommen, Silke Lauren; Zhao, Zhi; Segtnan, Sunniva; Stiansen-Sonerud, Tonje; Selvakumar, Joel Pradeepkumar & Havdal, Lise Beier [Show all 9 contributors for this article] (2024). Bulk RNA sequencing for analysis of post COVID-19 condition in adolescents and young adults. Journal of Translational Medicine. 22(1). doi: 10.1186/s12967-024-05117-7. Full text in Research Archive Show summary

  Background Post COVID19 condition (PCC) is a complication of SARSCOV2 infection and can lead to longterm disability. Methods The present study was designed to analyse the gene expression patterns of PCC through bulk RNA sequencing of whole blood and to explore the potential molecular mechanisms of PCC. Whole blood was collected from 80 participants enrolled in a prospective cohort study following SARSCoV2 infected and non-infected individuals for 6 months after recruitment and was used for bulk RNA sequencing. Identification of differentially expressed genes (DEG), pathway enrichment and immune cell deconvolution was performed to explore potential biological pathways involved in PCC. Results We have found 13 differentially expressed genes associated with PCC. Enriched pathways were related to interferon - signalling and antiviral immune processes. Conclusion The PCC transcriptome is characterized by a modest overexpression of interferon-stimulated genes, pointing to a subtle ongoing inflammatory response.

• Zhao, Zhi; Banterle, Marco; Lewin, Alex & Zucknick, Manuela (2023). Multivariate Bayesian structured variable selection for pharmacogenomic studies. Journal of the Royal Statistical Society, Series C (Applied Statistics). ISSN 0035-9254. 73(2), p. 420–443. doi: 10.1093/jrsssc/qlad102. Full text in Research Archive Show summary

  Abstract Cancer drug sensitivity screens combined with multi-omics characterisation of the cancer cells have become an important tool to determine the optimal treatment for each patient. We propose a multivariate Bayesian structured variable selection model for sparse identification of multi-omics features associated with multiple correlated drug responses. Our model uses known structure between drugs and their targeted genes via a Markov random field (MRF) prior in sparse seemingly unrelated regression. The use of MRF prior can improve the model performance compared to other common priors. The proposed model is applied to the Genomics of Drug Sensitivity in Cancer data.

• Wu, Yuwan; Wang, Zheng; Zhao, Zhi; Song, Xiuxia; Miao, Maohua & Zhang, Xi (2023). Bile acid metabolites in early pregnancy and risk of gestational diabetes mellitus: Results from a prospective cohort study. Diabetes, obesity and metabolism. ISSN 1462-8902. 25(8), p. 2255–2267. doi: 10.1111/dom.15104. Full text in Research Archive
• Zhao, Zhi; Zucknick, Manuela & Aittokallio, Tero Antero (2022). EnrichIntersect: an R package for custom set enrichment analysis and interactive visualization of intersecting sets. Bioinformatics Advances. 2(1). doi: 10.1093/bioadv/vbac073. Full text in Research Archive Show summary

  Abstract Summary Enrichment analysis has been widely used to study whether predefined sets of genes or other molecular features are over-represented in a ranked list associated with a disease or other phenotype. However, computational tools that perform enrichment analysis and visualization are usually limited to predefined sets available from public databases. To make such analyses more flexible, we introduce an R package, EnrichIntersect, which enables enrichment analyses among any ranked features and user-defined custom sets. For interactive visualization of multiple covariates, such as genes or other features, which are associated with multiple phenotypes and multiple sample groups, such as drug responses in various cancer types, EnrichIntersect illustrates all associations at a glance, hence explicitly indicating intersecting covariates between multiple phenotypic variables and between multiple sample groups. Availability and implementation The EnrichIntersect R package is available at https://CRAN.R-project.org/package=EnrichIntersect via an open-source MIT license. A package installation process is described on CRAN at https://cran.r-project.org/. A user-manual description of features and function calls can be found from the vignette of our package on CRAN.

• Zhao, Zhi; Wang, Shixiong; Zucknick, Manuela & Aittokallio, Tero Antero (2022). Tissue-specific identification of multi-omics features for pan-cancer drug response prediction. iScience. 25(8). doi: 10.1016/j.isci.2022.104767. Full text in Research Archive Show summary

  Current statistical models for drug response prediction and biomarker identification fall short in leveraging the shared and unique information from various cancer tissues and multi-omics profiles. We developed mix-lasso model that introduces an additional sample group penalty term to capture tissue-specific effects of features on pan-cancer response prediction. The mix-lasso model takes into account both the similarity between drug responses (i.e., multi-task learning), and the heterogeneity between multi-omics data (multi-modal learning). When applied to large-scale pharmacogenomics dataset from Cancer Therapeutics Response Portal, mix-lasso enabled accurate drug response predictions and identification of tissue-specific predictive features in the presence of various degrees of missing data, drug-drug correlations, and high-dimensional and correlated genomic and molecular features that often hinder the use of statistical approaches in drug response modeling. Compared to tree lasso model, mix-lasso identified a smaller number of tissue-specific features, hence making the model more interpretable and stable for drug discovery applications.

• Zhao, Zhi; Banterle, Marco; Bottolo, Leonardo; Richardson, Sylvia; Lewin, Alexandra & Zucknick, Manuela (2021). BayesSUR: An R package for high-dimensional multivariate Bayesian variable and covariance selection in linear regression. Journal of Statistical Software. 100(11), p. 1–32. doi: 10.18637/jss.v100.i11. Full text in Research Archive Show summary

  In molecular biology, advances in high-throughput technologies have made it possible to study complex multivariate phenotypes and their simultaneous associations with high-dimensional genomic and other omics data, a problem that can be studied with high-dimensional multi-response regression, where the response variables are potentially highly correlated. To this purpose, we recently introduced several multivariate Bayesian variable and covariance selection models, e.g., Bayesian estimation methods for sparse seemingly unrelated regression for variable and covariance selection. Several variable selection priors have been implemented in this context, in particular the hotspot detection prior for latent variable inclusion indicators, which results in sparse variable selection for associations between predictors and multiple phenotypes. We also propose an alternative, which uses a Markov random field (MRF) prior for incorporating prior knowledge about the dependence structure of the inclusion indicators. Inference of Bayesian seemingly unrelated regression (SUR) by Markov chain Monte Carlo methods is made computationally feasible by factorization of the covariance matrix amongst the response variables. In this paper we present BayesSUR, an R package, which allows the user to easily specify and run a range of different Bayesian SUR models, which have been implemented in C++ for computational efficiency. The R package allows the specification of the models in a modular way, where the user chooses the priors for variable selection and for covariance selection separately. We demonstrate the performance of sparse SUR models with the hotspot prior and spike-and-slab MRF prior on synthetic and real data sets representing eQTL or mQTL studies and in vitro anti-cancer drug screening studies as examples for typical applications.

• Mai, The Tien; Rønneberg, Leiv; Zhao, Zhi; Zucknick, Manuela & Corander, Jukka (2020). Learning Cancer Drug Sensitivities in Large-Scale Screens from Multi-omics Data with Local Low-Rank Structure. Lecture Notes in Computer Science (LNCS). ISSN 0302-9743. 12313, p. 67–79. doi: 10.1007/978-3-030-63061-4_7. Full text in Research Archive
• Gao, Yongxiang; Zhao, Zhi; Chen, Yimin; Mahara, Gehendra; Huang, Jialing & Lin, Zhuochen [Show all 7 contributors for this article] (2020). Automatic epileptic seizure classification in multichannel EEG time series with linear discriminant analysis. Technology and Health Care. ISSN 0928-7329. 28(1), p. 23–33. doi: 10.3233/THC-181548. Full text in Research Archive Show summary

  BACKGROUND: An electroencephalogram (EEG) is the most dominant method for detecting epileptic seizures. However, the existing techniques use single-channel EEGs from public databases and the sample size is small. OBJECTIVE: This study proposes a strategy to distinguish multichannel EEGs for health control, particularly the interictal and ictal multichannel EEGs of epileptic patients. METHOD: We calculated five features (variance, Pearson correlation coefficient, Hoeffding’s D measure, Shannon entropy, inter-quartile range), which are based on maximal overlap discrete wavelet transform. These features were then fed into linear discriminant analysis for classification purposes. Finally, the proposed method was tested on data on 34 healthy people, 21 interictal patients and 30 ictal patients taken from a hospital. RESULTS: Our experimental results show that the accuracy between healthy and epileptic seizures was 96.88% and the area under the curve (AUC) is 1. The accuracy between interictal and epileptic seizures was 94.12% and the AUC was 0.97. We also obtained an accuracy and AUC equal to 1 for discrimination of interictal EEGs from normal. Finally, we obtained an AUC of 0.83 and an accuracy of 85.88% for discrimination in these three classes. Therefore, our study achieves sufficient performance. CONCLUSIONS: Our proposed method can serve as an auxiliary tool for clinicians who wish to make clinical decisions and reduces the burden of detecting epileptic seizures.

• Zhao, Zhi & Zucknick, Manuela (2020). Structured penalized regression for drug sensitivity prediction. Journal of the Royal Statistical Society, Series C (Applied Statistics). ISSN 0035-9254. 69(3), p. 525–545. doi: 10.1111/rssc.12400. Full text in Research Archive Show summary

  Large‐scale in vitro drug sensitivity screens are an important tool in personalized oncology to predict the effectiveness of potential cancer drugs. The prediction of the sensitivity of cancer cell lines to a panel of drugs is a multivariate regression problem with high dimensional heterogeneous multiomics data as input data and with potentially strong correlations between the outcome variables which represent the sensitivity to the different drugs. We propose a joint penalized regression approach with structured penalty terms which enable us to utilize the correlation structure between drugs with group‐lasso‐type penalties and at the same time address the heterogeneity between ‘omics’ data sources by introducing data‐source‐specific penalty factors to penalize different data sources differently. By combining integrative penalty factors (IPFs) with the tree‐guided group lasso, we create a method called ‘IPF‐tree‐lasso’. We present a unified framework to transform more general IPF‐type methods to the original penalized method. Because the structured penalty terms have multiple parameters, we demonstrate how the interval search ‘Efficient parameter selection via global optimization’ algorithm can be used to optimize multiple penalty parameters efficiently. Simulation studies show that IPF‐tree‐lasso can improve the prediction performance compared with other lasso‐type methods, in particular for heterogeneous sources of data. Finally, we employ the new methods to analyse data from the ‘Genomics of drug sensitivity in cancer’ project.

• Pan, Hui; Shi, Xiaoshun; Xiao, Dakai; He, Jiaxi; Zhang, Yalei & Liang, Wenhua [Show all 11 contributors for this article] (2017). Nomogram prediction for the survival of the patients with small cell lung cancer. Journal of Thoracic Disease (JTD). ISSN 2072-1439. 9(3), p. 507–518. doi: 10.21037/jtd.2017.03.121. Full text in Research Archive
• Lin, Yi; Lin, Han; Lin, Qiuping; Zhang, Jinxin; Zhu, Ping & Lu, Yao [Show all 10 contributors for this article] (2016). A novel three-dimensional smile analysis based on dynamic evaluation of facial curve contour. Scientific Reports. 6(1). doi: 10.1038/srep22103. Full text in Research Archive

View all works in NVA

• Zhang, Jinxin; Zhao, Zhi; Xue, Yunlian; Chen, Zicong; Ma, Xinghua & Zhou, Qian (2017). Time series analysis. In Fang, Ji-Qian (Eds.), Handbook of Medical Statistics. World Scientific. ISSN 9789813148956. p. 269–299. doi: 10.1142/9789813148963_0009. Full text in Research Archive
• Zhao, Zhi (2020). Multivariate structured penalized and Bayesian regressions for pharmacogenomic screens. Universitetet i Oslo. Full text in Research Archive Show summary

  Pharmacogenomic screens for personalized cancer therapy are the focused biomedical application in this thesis. Due to the complex relationships between targeted cancer drugs and high-dimensional genomic predictors, we have developed penalized likelihood methods and Bayesian hierarchical models to capture the complex structures in the pharmacogenomic data and to predict drug sensitivity. The first part of the thesis proposed to address the correlations between drug sensitivity measures for multiple cancer drugs and the heterogeneity of multiple sources of genomic data in multivariate penalized likelihood methods with structured penalities. The proposed methods can improve the prediction performance of drug sensitivity. The second part of the thesis exploited Bayesian priors for the relationships between multiple drugs and relationships between drug sensitivity and the targeted pathways or genes of cancer drugs. Large pharmocogenomic screens may also include samples from multiple cancer tissue types. We employed random effects to address the sample heterogeneity in the proposed Bayesian model. The results have shown good structure recovery in the complex data and good prediction of responses by the new Bayesian models.

View all works in NVA