Federated learning framework integrating REFINED CNN and Deep Regression Forests.

Summary: Predictive learning from medical data incurs additional challenge due to concerns over privacy and security of personal data. Federated learning, intentionally structured to preserve high level of privacy, is emerging to be an attractive way to generate cross-silo predictions in medical scenarios. However, the impact of severe population-level heterogeneity on federated learners is not well explored. In this article, we propose a methodology to detect presence of population heterogeneity in federated settings and propose a solution to handle such heterogeneity by developing a federated version of Deep Regression Forests. Additionally, we demonstrate that the recently conceptualized REpresentation of Features as Images with NEighborhood Dependencies CNN framework can be combined with the proposed Federated Deep Regression Forests to provide improved performance as compared to existing approaches. Availability and implementation: The Python source code for reproducing the main results are available on GitHub: https://github.com/DanielNolte/FederatedDeepRegressionForests. Contact: ranadip.pal@ttu.edu. Supplementary information: Supplementary data are available at Bioinformatics Advances online.

Predicting binding affinities of emerging variants of SARS-CoV-2 using spike protein sequencing data: observations, caveats and recommendations.

Predicting protein properties from amino acid sequences is an important problem in biology and pharmacology. Protein-protein interactions among SARS-CoV-2 spike protein, human receptors and antibodies are key determinants of the potency of this virus and its ability to evade the human immune response. As a rapidly evolving virus, SARS-CoV-2 has already developed into many variants with considerable variation in virulence among these variants. Utilizing the proteomic data of SARS-CoV-2 to predict its viral characteristics will, therefore, greatly aid in disease control and prevention. In this paper, we review and compare recent successful prediction methods based on long short-term memory (LSTM), transformer, convolutional neural network (CNN) and a similarity-based topological regression (TR) model and offer recommendations about appropriate predictive methodology depending on the similarity between training and test datasets. We compare the effectiveness of these models in predicting the binding affinity and expression of SARS-CoV-2 spike protein sequences. We also explore how effective these predictive methods are when trained on laboratory-created data and are tasked with predicting the binding affinity of the in-the-wild SARS-CoV-2 spike protein sequences obtained from the GISAID datasets. We observe that TR is a better method when the sample size is small and test protein sequences are sufficiently similar to the training sequence. However, when the training sample size is sufficiently large and prediction requires extrapolation, LSTM embedding and CNN-based predictive model show superior performance.

Integrated multiomics analysis identifies molecular landscape perturbations during hyperammonemia in skeletal muscle and myotubes.

Ammonia is a cytotoxic molecule generated during normal cellular functions. Dysregulated ammonia metabolism, which is evident in many chronic diseases such as liver cirrhosis, heart failure, and chronic obstructive pulmonary disease, initiates a hyperammonemic stress response in tissues including skeletal muscle and in myotubes. Perturbations in levels of specific regulatory molecules have been reported, but the global responses to hyperammonemia are unclear. In this study, we used a multiomics approach to vertically integrate unbiased data generated using an assay for transposase-accessible chromatin with high-throughput sequencing, RNA-Seq, and proteomics. We then horizontally integrated these data across different models of hyperammonemia, including myotubes and mouse and human muscle tissues. Changes in chromatin accessibility and/or expression of genes resulted in distinct clusters of temporal molecular changes including transient, persistent, and delayed responses during hyperammonemia in myotubes. Known responses to hyperammonemia, including mitochondrial and oxidative dysfunction, protein homeostasis disruption, and oxidative stress pathway activation, were enriched in our datasets. During hyperammonemia, pathways that impact skeletal muscle structure and function that were consistently enriched were those that contribute to mitochondrial dysfunction, oxidative stress, and senescence. We made several novel observations, including an enrichment in antiapoptotic B-cell leukemia/lymphoma 2 family protein expression, increased calcium flux, and increased protein glycosylation in myotubes and muscle tissue upon hyperammonemia. Critical molecules in these pathways were validated experimentally. Human skeletal muscle from patients with cirrhosis displayed similar responses, establishing translational relevance. These data demonstrate complex molecular interactions during adaptive and maladaptive responses during the cellular stress response to hyperammonemia.

Investigation of REFINED CNN ensemble learning for anti-cancer drug sensitivity prediction.

MOTIVATION: Anti-cancer drug sensitivity prediction using deep learning models for individual cell line is a significant challenge in personalized medicine. Recently developed REFINED (REpresentation of Features as Images with NEighborhood Dependencies) CNN (Convolutional Neural Network)-based models have shown promising results in improving drug sensitivity prediction. The primary idea behind REFINED-CNN is representing high dimensional vectors as compact images with spatial correlations that can benefit from CNN architectures. However, the mapping from a high dimensional vector to a compact 2D image depends on the a priori choice of the distance metric and projection scheme with limited empirical procedures guiding these choices. RESULTS: In this article, we consider an ensemble of REFINED-CNN built under different choices of distance metrics and/or projection schemes that can improve upon a single projection based REFINED-CNN model. Results, illustrated using NCI60 and NCI-ALMANAC databases, demonstrate that the ensemble approaches can provide significant improvement in prediction performance as compared to individual models. We also develop the theoretical framework for combining different distance metrics to arrive at a single 2D mapping. Results demonstrated that distance-averaged REFINED-CNN produced comparable performance as obtained from stacking REFINED-CNN ensemble but with significantly lower computational cost. AVAILABILITY AND IMPLEMENTATION: The source code, scripts, and data used in the paper have been deposited in GitHub (https://github.com/omidbazgirTTU/IntegratedREFINED). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Representation of features as images with neighborhood dependencies for compatibility with convolutional neural networks.

Deep learning with Convolutional Neural Networks has shown great promise in image-based classification and enhancement but is often unsuitable for predictive modeling using features without spatial correlations. We present a feature representation approach termed REFINED (REpresentation of Features as Images with NEighborhood Dependencies) to arrange high-dimensional vectors in a compact image form conducible for CNN-based deep learning. We consider the similarities between features to generate a concise feature map in the form of a two-dimensional image by minimizing the pairwise distance values following a Bayesian Metric Multidimensional Scaling Approach. We hypothesize that this approach enables embedded feature extraction and, integrated with CNN-based deep learning, can boost the predictive accuracy. We illustrate the superior predictive capabilities of the proposed framework as compared to state-of-the-art methodologies in drug sensitivity prediction scenarios using synthetic datasets, drug chemical descriptors as predictors from NCI60, and both transcriptomic information and drug descriptors as predictors from GDSC.

Deep Functional and Molecular Characterization of a High-Risk Undifferentiated Pleomorphic Sarcoma.

Nonrhabdomyosarcoma soft-tissue sarcomas (STSs) are a class of 50+ cancers arising in muscle and soft tissues of children, adolescents, and adults. Rarity of each subtype often precludes subtype-specific preclinical research, leaving many STS patients with limited treatment options should frontline therapy be insufficient. When clinical options are exhausted, personalized therapy assignment approaches may help direct patient care. Here, we report the results of an adult female STS patient with relapsed undifferentiated pleomorphic sarcoma (UPS) who self-drove exploration of a wide array of personalized Clinical Laboratory Improvement Amendments (CLIAs) level and research-level diagnostics, including state of the art genomic, proteomic, ex vivo live cell chemosensitivity testing, a patient-derived xenograft model, and immunoscoring. Her therapeutic choices were also diverse, including neoadjuvant chemotherapy, radiation therapy, and surgeries. Adjuvant and recurrence strategies included off-label and natural medicines, several immunotherapies, and N-of-1 approaches. Identified treatment options, especially those validated during the in vivo study, were not introduced into the course of clinical treatment but did provide plausible treatment regimens based on FDA-approved clinical agents.

Evaluating the consistency of large-scale pharmacogenomic studies.

Recent years have seen an increase in the availability of pharmacogenomic databases such as Genomics of Drug Sensitivity in Cancer (GDSC) and Cancer Cell Line Encyclopedia (CCLE) that provide genomic and functional characterization information for multiple cell lines. Studies have alluded to the fact that specific characterizations may be inconsistent between different databases. Analysis of the potential discrepancies in the different databases is highly significant, as these sources are frequently used to analyze and validate methodologies for personalized cancer therapies. In this article, we review the recent developments in investigating the correspondence between different pharmacogenomics databases and discuss the potential factors that require attention when incorporating these sources in any modeling analysis. Furthermore, we explored the consistency among these databases using copulas that can capture nonlinear dependencies between two sets of data.

Probabilistic modeling of personalized drug combinations from integrated chemical screen and molecular data in sarcoma.

BACKGROUND: Cancer patients with advanced disease routinely exhaust available clinical regimens and lack actionable genomic medicine results, leaving a large patient population without effective treatments options when their disease inevitably progresses. To address the unmet clinical need for evidence-based therapy assignment when standard clinical approaches have failed, we have developed a probabilistic computational modeling approach which integrates molecular sequencing data with functional assay data to develop patient-specific combination cancer treatments. METHODS: Tissue taken from a murine model of alveolar rhabdomyosarcoma was used to perform single agent drug screening and DNA/RNA sequencing experiments; results integrated via our computational modeling approach identified a synergistic personalized two-drug combination. Cells derived from the primary murine tumor were allografted into mouse models and used to validate the personalized two-drug combination. Computational modeling of single agent drug screening and RNA sequencing of multiple heterogenous sites from a single patient's epithelioid sarcoma identified a personalized two-drug combination effective across all tumor regions. The heterogeneity-consensus combination was validated in a xenograft model derived from the patient's primary tumor. Cell cultures derived from human and canine undifferentiated pleomorphic sarcoma were assayed by drug screen; computational modeling identified a resistance-abrogating two-drug combination common to both cell cultures. This combination was validated in vitro via a cell regrowth assay. RESULTS: Our computational modeling approach addresses three major challenges in personalized cancer therapy: synergistic drug combination predictions (validated in vitro and in vivo in a genetically engineered murine cancer model), identification of unifying therapeutic targets to overcome intra-tumor heterogeneity (validated in vivo in a human cancer xenograft), and mitigation of cancer cell resistance and rewiring mechanisms (validated in vitro in a human and canine cancer model). CONCLUSIONS: These proof-of-concept studies support the use of an integrative functional approach to personalized combination therapy prediction for the population of high-risk cancer patients lacking viable clinical options and without actionable DNA sequencing-based therapy.

Recursive model for dose-time responses in pharmacological studies.

BACKGROUND: Clinical studies often track dose-response curves of subjects over time. One can easily model the dose-response curve at each time point with Hill equation, but such a model fails to capture the temporal evolution of the curves. On the other hand, one can use Gompertz equation to model the temporal behaviors at each dose without capturing the evolution of time curves across dosage. RESULTS: In this article, we propose a parametric model for dose-time responses that follows Gompertz law in time and Hill equation across dose approximately. We derive a recursion relation for dose-response curves over time capturing the temporal evolution and then specify a regression model connecting the parameters controlling the dose-time responses with individual level proteomic data. The resultant joint model allows us to predict the dose-response curves over time for new individuals. CONCLUSION: We have compared the efficacy of our proposed Recursive Hybrid model with individual dose-response predictive models at desired time points. We note that our proposed model exhibits a superior performance compared to the individual ones for both synthetic data and actual pharmacological data. For the desired dose-time varying genetic characterization and drug response values, we have used the HMS-LINCS database and demonstrated the effectiveness of our model for all available anticancer compounds.

Preclinical rationale for entinostat in embryonal rhabdomyosarcoma.

BACKGROUND: Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma in the pediatric cancer population. Survival among metastatic RMS patients has remained dismal yet unimproved for years. We previously identified the class I-specific histone deacetylase inhibitor, entinostat (ENT), as a pharmacological agent that transcriptionally suppresses the PAX3:FOXO1 tumor-initiating fusion gene found in alveolar rhabdomyosarcoma (aRMS), and we further investigated the mechanism by which ENT suppresses PAX3:FOXO1 oncogene and demonstrated the preclinical efficacy of ENT in RMS orthotopic allograft and patient-derived xenograft (PDX) models. In this study, we investigated whether ENT also has antitumor activity in fusion-negative eRMS orthotopic allografts and PDX models either as a single agent or in combination with vincristine (VCR). METHODS: We tested the efficacy of ENT and VCR as single agents and in combination in orthotopic allograft and PDX mouse models of eRMS. We then performed CRISPR screening to identify which HDAC among the class I HDACs is responsible for tumor growth inhibition in eRMS. To analyze whether ENT treatment as a single agent or in combination with VCR induces myogenic differentiation, we performed hematoxylin and eosin (H&E) staining in tumors. RESULTS: ENT in combination with the chemotherapy VCR has synergistic antitumor activity in a subset of fusion-negative eRMS in orthotopic "allografts," although PDX mouse models were too hypersensitive to the VCR dose used to detect synergy. Mechanistic studies involving CRISPR suggest that HDAC3 inhibition is the primary mechanism of cell-autonomous cytoreduction in eRMS. Following cytoreduction in vivo, residual tumor cells in the allograft models treated with chemotherapy undergo a dramatic, entinostat-induced (70-100%) conversion to non-proliferative rhabdomyoblasts. CONCLUSION: Our results suggest that the targeting class I HDACs may provide a therapeutic benefit for selected patients with eRMS. ENT's preclinical in vivo efficacy makes ENT a rational drug candidate in a phase II clinical trial for eRMS.

Functional random forest with applications in dose-response predictions.

Drug sensitivity prediction for individual tumors is a significant challenge in personalized medicine. Current modeling approaches consider prediction of a single metric of the drug response curve such as AUC or IC50. However, the single summary metric of a dose-response curve fails to provide the entire drug sensitivity profile which can be used to design the optimal dose for a patient. In this article, we assess the problem of predicting the complete dose-response curve based on genetic characterizations. We propose an enhancement to the popular ensemble-based Random Forests approach that can directly predict the entire functional profile of a dose-response curve rather than a single summary metric. We design functional regression trees with node costs modified based on dose/response region dependence methodologies and response distribution based approaches. Our results relative to large pharmacological databases such as CCLE and GDSC show a higher accuracy in predicting dose-response curves of the proposed functional framework in contrast to univariate or multivariate Random Forest predicting sensitivities at different dose levels. Furthermore, we also considered the problem of predicting functional responses from functional predictors i.e., estimating the dose-response curves with a model built on dose-dependent expression data. The superior performance of Functional Random Forest using functional data as compared to existing approaches have been shown using the HMS-LINCS dataset. In summary, Functional Random Forest presents an enhanced predictive modeling framework to predict the entire functional response profile considering both static and functional predictors instead of predicting the summary metrics of the response curves.

Sstack: an R package for stacking with applications to scenarios involving sequential addition of samples and features.

SUMMARY: Biological processes are characterized by a variety of different genomic feature sets. However, often times when building models, portions of these features are missing for a subset of the dataset. We provide a modeling framework to effectively integrate this type of heterogeneous data to improve prediction accuracy. To test our methodology, we have stacked data from the Cancer Cell Line Encyclopedia to increase the accuracy of drug sensitivity prediction. The package addresses the dynamic regime of information integration involving sequential addition of features and samples. AVAILABILITY AND IMPLEMENTATION: The framework has been implemented as a R package Sstack, which can be downloaded from https://cran.r-project.org/web/packages/Sstack/index.html, where further explanation of the package is available. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Predictive Modeling of Anti-Cancer Drug Sensitivity from Genetic Characterizations.

Accurately predicting sensitivity of tumor cells to anti-cancer drugs based on genetic characterizations is a significant challenge for personalized cancer therapy. This chapter provides a computational procedure to design predictive models from individual genomic characterizations and combine them to arrive at an integrated predictive model. Integrated modeling employs the complementary information from heterogeneous genetic characterizations to improve the prediction error as well as lowering the error confidence interval.

Application of transfer learning for cancer drug sensitivity prediction.

BACKGROUND: In precision medicine, scarcity of suitable biological data often hinders the design of an appropriate predictive model. In this regard, large scale pharmacogenomics studies, like CCLE and GDSC hold the promise to mitigate the issue. However, one cannot directly employ data from multiple sources together due to the existing distribution shift in data. One way to solve this problem is to utilize the transfer learning methodologies tailored to fit in this specific context. RESULTS: In this paper, we present two novel approaches for incorporating information from a secondary database for improving the prediction in a target database. The first approach is based on latent variable cost optimization and the second approach considers polynomial mapping between the two databases. Utilizing CCLE and GDSC databases, we illustrate that the proposed approaches accomplish a better prediction of drug sensitivities for different scenarios as compared to the existing approaches. CONCLUSION: We have compared the performance of the proposed predictive models with database-specific individual models as well as existing transfer learning approaches. We note that our proposed approaches exhibit superior performance compared to the abovementioned alternative techniques for predicting sensitivity for different anti-cancer compounds, particularly the nonlinear mapping model shows the best overall performance.

The HDAC3-SMARCA4-miR-27a axis promotes expression of the PAX3:FOXO1 fusion oncogene in rhabdomyosarcoma.

Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma of childhood with an unmet clinical need for decades. A single oncogenic fusion gene is associated with treatment resistance and a 40 to 45% decrease in overall survival. We previously showed that expression of this PAX3:FOXO1 fusion oncogene in alveolar RMS (aRMS) mediates tolerance to chemotherapy and radiotherapy and that the class I-specific histone deacetylase (HDAC) inhibitor entinostat reduces PAX3:FOXO1 protein abundance. Here, we established the antitumor efficacy of entinostat with chemotherapy in various preclinical cell and mouse models and found that HDAC3 inhibition was the primary mechanism of entinostat-induced suppression of PAX3:FOXO1 abundance. HDAC3 inhibition by entinostat decreased the activity of the chromatin remodeling enzyme SMARCA4, which, in turn, derepressed the microRNA miR-27a. This reexpression of miR-27a led to PAX3:FOXO1 mRNA destabilization and chemotherapy sensitization in aRMS cells in culture and in vivo. Furthermore, a phase 1 clinical trial (ADVL1513) has shown that entinostat is tolerable in children with relapsed or refractory solid tumors and is planned for phase 1B cohort expansion or phase 2 clinical trials. Together, these results implicate an HDAC3-SMARCA4-miR-27a-PAX3:FOXO1 circuit as a driver of chemoresistant aRMS and suggest that targeting this pathway with entinostat may be therapeutically effective in patients.

Dimensionality Reduction based Transfer Learning applied to Pharmacogenomics Databases.

Recent years have observed a number of Pharmacogenomics databases being published that enable testing of various predictive modeling techniques for personalized therapy applications. However, the consistencies between the databases are usually limited in spite of having significant number of common cell lines and drugs. In this article, we consider the problem of whether we can use the model learned from one secondary database to improve the prediction for the other target database. We illustrate using two pharmacogenomics databases that representing the databases using common basis vectors can improve prediction performance as compared to the naive application of a model trained on one database to another. We also elucidate the robustness of using PCA based basis vectors for scenarios with low correlated input features.

Investigation of model stacking for drug sensitivity prediction.

BACKGROUND: A significant problem in precision medicine is the prediction of drug sensitivity for individual cancer cell lines. Predictive models such as Random Forests have shown promising performance while predicting from individual genomic features such as gene expressions. However, accessibility of various other forms of data types including information on multiple tested drugs necessitates the examination of designing predictive models incorporating the various data types. RESULTS: We explore the predictive performance of model stacking and the effect of stacking on the predictive bias and squared error. In addition we discuss the analytical underpinnings supporting the advantages of stacking in reducing squared error and inherent bias of random forests in prediction of outliers. The framework is tested on a setup including gene expression, drug target, physical properties and drug response information for a set of drugs and cell lines. CONCLUSION: The performance of individual and stacked models are compared. We note that stacking models built on two heterogeneous datasets provide superior performance to stacking different models built on the same dataset. It is also noted that stacking provides a noticeable reduction in the bias of our predictors when the dominant eigenvalue of the principle axis of variation in the residuals is significantly higher than the remaining eigenvalues.