IU Indianapolis ScholarWorks :: Browsing by Author "Johnson, Travis S."

Browsing by Author "Johnson, Travis S."

Now showing 1 - 10 of 12

BERMUDA: a novel deep transfer learning method for single-cell RNA sequencing batch correction reveals hidden high-resolution cellular subtypes
(BioMed Central, 2019-08-12) Wang, Tongxin; Johnson, Travis S.; Shao, Wei; Lu, Zixiao; Helm, Bryan R.; Zhang, Jie; Huang, Kun; Medical and Molecular Genetics, School of Medicine
To fully utilize the power of single-cell RNA sequencing (scRNA-seq) technologies for identifying cell lineages and bona fide transcriptional signals, it is necessary to combine data from multiple experiments. We present BERMUDA (Batch Effect ReMoval Using Deep Autoencoders), a novel transfer-learning-based method for batch effect correction in scRNA-seq data. BERMUDA effectively combines different batches of scRNA-seq data with vastly different cell population compositions and amplifies biological signals by transferring information among batches. We demonstrate that BERMUDA outperforms existing methods for removing batch effects and distinguishing cell types in multiple simulated and real scRNA-seq datasets.
Combinatorial analyses reveal cellular composition changes have different impacts on transcriptomic changes of cell type specific genes in Alzheimer’s Disease
(Springer Nature, 2021-01-11) Johnson, Travis S.; Xiang, Shunian; Dong, Tianhan; Huang, Zhi; Cheng, Michael; Wang, Tianfu; Yang, Kai; Ni, Dong; Huang, Kun; Zhang, Jie; Biostatistics, School of Public Health
Alzheimer’s disease (AD) brains are characterized by progressive neuron loss and gliosis. Previous studies of gene expression using bulk tissue samples often fail to consider changes in cell-type composition when comparing AD versus control, which can lead to differences in expression levels that are not due to transcriptional regulation. We mined five large transcriptomic AD datasets for conserved gene co-expression module, then analyzed differential expression and differential co-expression within the modules between AD samples and controls. We performed cell-type deconvolution analysis to determine whether the observed differential expression was due to changes in cell-type proportions in the samples or to transcriptional regulation. Our findings were validated using four additional datasets. We discovered that the increased expression of microglia modules in the AD samples can be explained by increased microglia proportions in the AD samples. In contrast, decreased expression and perturbed co-expression within neuron modules in the AD samples was likely due in part to altered regulation of neuronal pathways. Several transcription factors that are differentially expressed in AD might account for such altered gene regulation. Similarly, changes in gene expression and co-expression within astrocyte modules could be attributed to combined effects of astrogliosis and astrocyte gene activation. Gene expression in the astrocyte modules was also strongly correlated with clinicopathological biomarkers. Through this work, we demonstrated that combinatorial analysis can delineate the origins of transcriptomic changes in bulk tissue data and shed light on key genes and pathways involved in AD.
Deep learning-based cancer survival prognosis from RNA-seq data: approaches and evaluations
(BMC, 2020) Huang, Zhi; Johnson, Travis S.; Han, Zhi; Helm, Bryan; Cao, Sha; Zhang, Chi; Salama, Paul; Rizkalla, Maher; Yu, Christina Y.; Cheng, Jun; Xiang, Shunian; Zhan, Xiaohui; Zhang, Jie; Huang, Kun; Medicine, School of Medicine
Background: Recent advances in kernel-based Deep Learning models have introduced a new era in medical research. Originally designed for pattern recognition and image processing, Deep Learning models are now applied to survival prognosis of cancer patients. Specifically, Deep Learning versions of the Cox proportional hazards models are trained with transcriptomic data to predict survival outcomes in cancer patients. Methods: In this study, a broad analysis was performed on TCGA cancers using a variety of Deep Learning-based models, including Cox-nnet, DeepSurv, and a method proposed by our group named AECOX (AutoEncoder with Cox regression network). Concordance index and p-value of the log-rank test are used to evaluate the model performances. Results: All models show competitive results across 12 cancer types. The last hidden layers of the Deep Learning approaches are lower dimensional representations of the input data that can be used for feature reduction and visualization. Furthermore, the prognosis performances reveal a negative correlation between model accuracy, overall survival time statistics, and tumor mutation burden (TMB), suggesting an association among overall survival time, TMB, and prognosis prediction accuracy. Conclusions: Deep Learning based algorithms demonstrate superior performances than traditional machine learning based models. The cancer prognosis results measured in concordance index are indistinguishable across models while are highly variable across cancers. These findings shedding some light into the relationships between patient characteristics and survival learnability on a pan-cancer level.
Gene Co-expression Network and Copy Number Variation Analyses Identify Transcription Factors Associated With Multiple Myeloma Progression
(Frontiers, 2019-05-17) Yu, Christina Y.; Xiang, Shunian; Huang, Zhi; Johnson, Travis S.; Zhan, Xiaohui; Han, Zhi; Abu Zaid, Mohammad; Huang, Kun; Medicine, School of Medicine
Multiple myeloma (MM) has two clinical precursor stages of disease: monoclonal gammopathy of undetermined significance (MGUS) and smoldering multiple myeloma (SMM). However, the mechanism of progression is not well understood. Because gene co-expression network analysis is a well-known method for discovering new gene functions and regulatory relationships, we utilized this framework to conduct differential co-expression analysis to identify interesting transcription factors (TFs) in two publicly available datasets. We then used copy number variation (CNV) data from a third public dataset to validate these TFs. First, we identified co-expressed gene modules in two publicly available datasets each containing three conditions: normal, MGUS, and SMM. These modules were assessed for condition-specific gene expression, and then enrichment analysis was conducted on condition-specific modules to identify their biological function and upstream TFs. TFs were assessed for differential gene expression between normal and MM precursors, then validated with CNV analysis to identify candidate genes. Functional enrichment analysis reaffirmed known functional categories in MM pathology, the main one relating to immune function. Enrichment analysis revealed a handful of differentially expressed TFs between normal and either MGUS or SMM in gene expression and/or CNV. Overall, we identified four genes of interest (MAX, TCF4, ZNF148, and ZNF281) that aid in our understanding of MM initiation and progression.
LAmbDA: label ambiguous domain adaptation dataset integration reduces batch effects and improves subtype detection
(Oxford Academic, 2019-04) Johnson, Travis S.; Wang, Tongxin; Huang, Zhi; Yu, Christina Y.; Wu, Yi; Han, Yatong; Zhang, Yan; Huang, Kun; Zhang, Jie; Medicine, School of Medicine
Motivation Rapid advances in single cell RNA sequencing (scRNA-seq) have produced higher-resolution cellular subtypes in multiple tissues and species. Methods are increasingly needed across datasets and species to (i) remove systematic biases, (ii) model multiple datasets with ambiguous labels and (iii) classify cells and map cell type labels. However, most methods only address one of these problems on broad cell types or simulated data using a single model type. It is also important to address higher-resolution cellular subtypes, subtype labels from multiple datasets, models trained on multiple datasets simultaneously and generalizability beyond a single model type. Results We developed a species- and dataset-independent transfer learning framework (LAmbDA) to train models on multiple datasets (even from different species) and applied our framework on simulated, pancreas and brain scRNA-seq experiments. These models mapped corresponding cell types between datasets with inconsistent cell subtype labels while simultaneously reducing batch effects. We achieved high accuracy in labeling cellular subtypes (weighted accuracy simulated 1 datasets: 90%; simulated 2 datasets: 94%; pancreas datasets: 88% and brain datasets: 66%) using LAmbDA Feedforward 1 Layer Neural Network with bagging. This method achieved higher weighted accuracy in labeling cellular subtypes than two other state-of-the-art methods, scmap and CaSTLe in brain (66% versus 60% and 32%). Furthermore, it achieved better performance in correctly predicting ambiguous cellular subtype labels across datasets in 88% of test cases compared with CaSTLe (63%), scmap (50%) and MetaNeighbor (50%). LAmbDA is model- and dataset-independent and generalizable to diverse data types representing an advance in biocomputing.
Network analysis of pseudogene-gene relationships: from pseudogene evolution to their functional potentials
(Pacific Symposium on Biocomputing, 2018) Johnson, Travis S.; Li, Sihong; Kho, Jonathan R.; Huang, Kun; Zhang, Yan; Medicine, School of Medicine
Pseudogenes are fossil relatives of genes. Pseudogenes have long been thought of as "junk DNAs", since they do not code proteins in normal tissues. Although most of the human pseudogenes do not have noticeable functions, ∼20% of them exhibit transcriptional activity. There has been evidence showing that some pseudogenes adopted functions as lncRNAs and work as regulators of gene expression. Furthermore, pseudogenes can even be "reactivated" in some conditions, such as cancer initiation. Some pseudogenes are transcribed in specific cancer types, and some are even translated into proteins as observed in several cancer cell lines. All the above have shown that pseudogenes could have functional roles or potentials in the genome. Evaluating the relationships between pseudogenes and their gene counterparts could help us reveal the evolutionary path of pseudogenes and associate pseudogenes with functional potentials. It also provides an insight into the regulatory networks involving pseudogenes with transcriptional and even translational activities.In this study, we develop a novel approach integrating graph analysis, sequence alignment and functional analysis to evaluate pseudogene-gene relationships, and apply it to human gene homologs and pseudogenes. We generated a comprehensive set of 445 pseudogene-gene (PGG) families from the original 3,281 gene families (13.56%). Of these 438 (98.4% PGG, 13.3% total) were non-trivial (containing more than one pseudogene). Each PGG family contains multiple genes and pseudogenes with high sequence similarity. For each family, we generate a sequence alignment network and phylogenetic trees recapitulating the evolutionary paths. We find evidence supporting the evolution history of olfactory family (both genes and pseudogenes) in human, which also supports the validity of our analysis method. Next, we evaluate these networks in respect to the gene ontology from which we identify functions enriched in these pseudogene-gene families and infer functional impact of pseudogenes involved in the networks. This demonstrates the application of our PGG network database in the study of pseudogene function in disease context.
A protocol to evaluate RNA sequencing normalization methods
(BMC, 2019-12-20) Abrams, Zachary B.; Johnson, Travis S.; Huang, Kun; Payne, Philip R. O.; Coombes, Kevin; Medicine, School of Medicine
Background RNA sequencing technologies have allowed researchers to gain a better understanding of how the transcriptome affects disease. However, sequencing technologies often unintentionally introduce experimental error into RNA sequencing data. To counteract this, normalization methods are standardly applied with the intent of reducing the non-biologically derived variability inherent in transcriptomic measurements. However, the comparative efficacy of the various normalization techniques has not been tested in a standardized manner. Here we propose tests that evaluate numerous normalization techniques and applied them to a large-scale standard data set. These tests comprise a protocol that allows researchers to measure the amount of non-biological variability which is present in any data set after normalization has been performed, a crucial step to assessing the biological validity of data following normalization. Results In this study we present two tests to assess the validity of normalization methods applied to a large-scale data set collected for systematic evaluation purposes. We tested various RNASeq normalization procedures and concluded that transcripts per million (TPM) was the best performing normalization method based on its preservation of biological signal as compared to the other methods tested. Conclusion Normalization is of vital importance to accurately interpret the results of genomic and transcriptomic experiments. More work, however, needs to be performed to optimize normalization methods for RNASeq data. The present effort helps pave the way for more systematic evaluations of normalization methods across different platforms. With our proposed schema researchers can evaluate their own or future normalization methods to further improve the field of RNASeq normalization.
PseudoFuN: Deriving functional potentials of pseudogenes from integrative relationships with genes and microRNAs across 32 cancers
(Oxford University Press, 2019-05) Johnson, Travis S.; Li, Sihong; Franz, Eric; Huang, Zhi; Dan Li, Shuyu; Campbell, Moray J.; Huang, Kun; Zhang, Yan; Medicine, School of Medicine
BACKGROUND: Long thought "relics" of evolution, not until recently have pseudogenes been of medical interest regarding regulation in cancer. Often, these regulatory roles are a direct by-product of their close sequence homology to protein-coding genes. Novel pseudogene-gene (PGG) functional associations can be identified through the integration of biomedical data, such as sequence homology, functional pathways, gene expression, pseudogene expression, and microRNA expression. However, not all of the information has been integrated, and almost all previous pseudogene studies relied on 1:1 pseudogene-parent gene relationships without leveraging other homologous genes/pseudogenes. RESULTS: We produce PGG families that expand beyond the current 1:1 paradigm. First, we construct expansive PGG databases by (i) CUDAlign graphics processing unit (GPU) accelerated local alignment of all pseudogenes to gene families (totaling 1.6 billion individual local alignments and >40,000 GPU hours) and (ii) BLAST-based assignment of pseudogenes to gene families. Second, we create an open-source web application (PseudoFuN [Pseudogene Functional Networks]) to search for integrative functional relationships of sequence homology, microRNA expression, gene expression, pseudogene expression, and gene ontology. We produce four "flavors" of CUDAlign-based databases (>462,000,000 PGG pairwise alignments and 133,770 PGG families) that can be queried and downloaded using PseudoFuN. These databases are consistent with previous 1:1 PGG annotation and also are much more powerful including millions of de novo PGG associations. For example, we find multiple known (e.g., miR-20a-PTEN-PTENP1) and novel (e.g., miR-375-SOX15-PPP4R1L) microRNA-gene-pseudogene associations in prostate cancer. PseudoFuN provides a "one stop shop" for identifying and visualizing thousands of potential regulatory relationships related to pseudogenes in The Cancer Genome Atlas cancers. CONCLUSIONS: Thousands of new PGG associations can be explored in the context of microRNA-gene-pseudogene co-expression and differential expression with a simple-to-use online tool by bioinformaticians and oncologists alike.
Pseudogene-gene functional networks are prognostic of patient survival in breast cancer
(BMC, 2020) Smerekanych, Sasha; Johnson, Travis S.; Huang, Kun; Zhang, Yan; Medicine, School of Medicine
Background: Given the vast range of molecular mechanisms giving rise to breast cancer, it is unlikely universal cures exist. However, by providing a more precise prognosis for breast cancer patients through integrative models, treatments can become more individualized, resulting in more successful outcomes. Specifically, we combine gene expression, pseudogene expression, miRNA expression, clinical factors, and pseudogene-gene functional networks to generate these models for breast cancer prognostics. Establishing a LASSO-generated molecular gene signature revealed that the increased expression of genes STXBP5, GALP and LOC387646 indicate a poor prognosis for a breast cancer patient. We also found that increased CTSLP8 and RPS10P20 and decreased HLA-K pseudogene expression indicate poor prognosis for a patient. Perhaps most importantly we identified a pseudogene-gene interaction, GPS2-GPS2P1 (improved prognosis) that is prognostic where neither the gene nor pseudogene alone is prognostic of survival. Besides, miR-3923 was predicted to target GPS2 using miRanda, PicTar, and TargetScan, which imply modules of gene-pseudogene-miRNAs that are potentially functionally related to patient survival. Results: In our LASSO-based model, we take into account features including pseudogenes, genes and candidate pseudogene-gene interactions. Key biomarkers were identified from the features. The identification of key biomarkers in combination with significant clinical factors (such as stage and radiation therapy status) should be considered as well, enabling a specific prognostic prediction and future treatment plan for an individual patient. Here we used our PseudoFuN web application to identify the candidate pseudogene-gene interactions as candidate features in our integrative models. We further identified potential miRNAs targeting those features in our models using PseudoFuN as well. From this study, we present an interpretable survival model based on LASSO and decision trees, we also provide a novel feature set which includes pseudogene-gene interaction terms that have been ignored by previous prognostic models. We find that some interaction terms for pseudogenes and genes are significantly prognostic of survival. These interactions are cross-over interactions, where the impact of the gene expression on survival changes with pseudogene expression and vice versa. These may imply more complicated regulation mechanisms than previously understood. Conclusions: We recommend these novel feature sets be considered when training other types of prognostic models as well, which may provide more comprehensive insights into personalized treatment decisions.
SALMON: Survival Analysis Learning With Multi-Omics Neural Networks on Breast Cancer
(Frontiers Media, 2019-03-08) Huang, Zhi; Zhan, Xiaohui; Xiang, Shunian; Johnson, Travis S.; Helm, Bryan; Yu, Christina Y.; Zhang, Jie; Salama, Paul; Rizkalla, Maher; Han, Zhi; Huang, Kun; Department of Medicine, Indiana University School of Medicine
Improved cancer prognosis is a central goal for precision health medicine. Though many models can predict differential survival from data, there is a strong need for sophisticated algorithms that can aggregate and filter relevant predictors from increasingly complex data inputs. In turn, these models should provide deeper insight into which types of data are most relevant to improve prognosis. Deep Learning-based neural networks offer a potential solution for both problems because they are highly flexible and account for data complexity in a non-linear fashion. In this study, we implement Deep Learning-based networks to determine how gene expression data predicts Cox regression survival in breast cancer. We accomplish this through an algorithm called SALMON (Survival Analysis Learning with Multi-Omics Neural Networks), which aggregates and simplifies gene expression data and cancer biomarkers to enable prognosis prediction. The results revealed improved performance when more omics data were used in model construction. Rather than use raw gene expression values as model inputs, we innovatively use eigengene modules from the result of gene co-expression network analysis. The corresponding high impact co-expression modules and other omics data are identified by feature selection technique, then examined by conducting enrichment analysis and exploiting biological functions, escalated the interpretation of input feature from gene level to co-expression modules level. Our study shows the feasibility of discovering breast cancer related co-expression modules, sketch a blueprint of future endeavors on Deep Learning-based survival analysis. SALMON source code is available at https://github.com/huangzhii/SALMON/.

Browsing by Author "Johnson, Travis S."

Results Per Page

Sort Options