TransCell: In Silico Characterization of Genomic Landscape and Cellular Responses by Deep Transfer Learning.

Gene expression profiling of new or modified cell lines becomes routine today; however, obtaining comprehensive molecular characterization and cellular responses for a variety of cell lines, including those derived from underrepresented groups, is not trivial when resources are minimal. Using gene expression to predict other measurements has been actively explored; however, systematic investigation of its predictive power in various measurements has not been well studied. Here, we evaluated commonly used machine learning methods and presented TransCell, a two-step deep transfer learning framework that utilized the knowledge derived from pan-cancer tumor samples to predict molecular features and responses. Among these models, TransCell had the best performance in predicting metabolite, gene effect score (or genetic dependency), and drug sensitivity, and had comparable performance in predicting mutation, copy number variation, and protein expression. Notably, TransCell improved the performance by over 50% in drug sensitivity prediction and achieved a correlation of 0.7 in gene effect score prediction. Furthermore, predicted drug sensitivities revealed potential repurposing candidates for new 100 pediatric cancer cell lines, and predicted gene effect scores reflected BRAF resistance in melanoma cell lines. Together, we investigated the predictive power of gene expression in six molecular measurement types and developed a web portal (http://apps.octad.org/transcell/) that enables the prediction of 352,000 genomic and cellular response features solely from gene expression profiles.

Large-Scale Information Retrieval and Correction of Noisy Pharmacogenomic Datasets through Residual Thresholded Deep Matrix Factorization.

Pharmacogenomics studies are attracting an increasing amount of interest from researchers in precision medicine. The advances in high-throughput experiments and multiplexed approaches allow the large-scale quantification of drug sensitivities in molecularly characterized cancer cell lines (CCLs), resulting in a number of open drug sensitivity datasets for drug biomarker discovery. However, a significant inconsistency in drug sensitivity values among these datasets has been noted. Such inconsistency indicates the presence of substantial noise, subsequently hindering downstream analyses. To address the noise in drug sensitivity data, we introduce a robust and scalable deep learning framework, Residual Thresholded Deep Matrix Factorization (RT-DMF). This method takes a single drug sensitivity data matrix as its sole input and outputs a corrected and imputed matrix. Deep Matrix Factorization (DMF) excels at uncovering subtle patterns, due to its minimal reliance on data structure assumptions. This attribute significantly boosts DMF's ability to identify complex hidden patterns among nuisance effects in the data, thereby facilitating the detection of signals that are therapeutically relevant. Furthermore, RT-DMF incorporates an iterative residual thresholding (RT) procedure, which plays a crucial role in retaining signals more likely to hold therapeutic importance. Validation using simulated datasets and real pharmacogenomics datasets demonstrates the effectiveness of our approach in correcting noise and imputing missing data in drug sensitivity datasets (open source package available at https://github.com/tomwhoooo/rtdmf).

Systems Drug Discovery for Diffuse Large B Cell Lymphoma Based on Pathogenic Molecular Mechanism via Big Data Mining and Deep Learning Method.

Diffuse large B cell lymphoma (DLBCL) is an aggressive heterogeneous disease. The most common subtypes of DLBCL include germinal center b-cell (GCB) type and activated b-cell (ABC) type. To learn more about the pathogenesis of two DLBCL subtypes (i.e., DLBCL ABC and DLBCL GCB), we firstly construct a candidate genome-wide genetic and epigenetic network (GWGEN) by big database mining. With the help of two DLBCL subtypes' genome-wide microarray data, we identify their real GWGENs via system identification and model order selection approaches. Afterword, the core GWGENs of two DLBCL subtypes could be extracted from real GWGENs by principal network projection (PNP) method. By comparing core signaling pathways and investigating pathogenic mechanisms, we are able to identify pathogenic biomarkers as drug targets for DLBCL ABC and DLBCL GCD, respectively. Furthermore, we do drug discovery considering drug-target interaction ability, drug regulation ability, and drug toxicity. Among them, a deep neural network (DNN)-based drug-target interaction (DTI) model is trained in advance to predict potential drug candidates holding higher probability to interact with identified biomarkers. Consequently, two drug combinations are proposed to alleviate DLBCL ABC and DLBCL GCB, respectively.

Investigating the Role of Obesity in Prostate Cancer and Identifying Biomarkers for Drug Discovery: Systems Biology and Deep Learning Approaches.

Prostate cancer (PCa) is the second most frequently diagnosed cancer for men and is viewed as the fifth leading cause of death worldwide. The body mass index (BMI) is taken as a vital criterion to elucidate the association between obesity and PCa. In this study, systematic methods are employed to investigate how obesity influences the noncutaneous malignancies of PCa. By comparing the core signaling pathways of lean and obese patients with PCa, we are able to investigate the relationships between obesity and pathogenic mechanisms and identify significant biomarkers as drug targets for drug discovery. Regarding drug design specifications, we take drug-target interaction, drug regulation ability, and drug toxicity into account. One deep neural network (DNN)-based drug-target interaction (DTI) model is trained in advance for predicting drug candidates based on the identified biomarkers. In terms of the application of the DNN-based DTI model and the consideration of drug design specifications, we suggest two potential multiple-molecule drugs to prevent PCa (covering lean and obese PCa) and obesity-specific PCa, respectively. The proposed multiple-molecule drugs (apigenin, digoxin, and orlistat) not only help to prevent PCa, suppressing malignant metastasis, but also result in lower production of fatty acids and cholesterol, especially for obesity-specific PCa.

Systems Medicine Design based on Systems Biology Approaches and Deep Neural Network for Gastric Cancer.

Gastric cancer (GC) is the third leading cause of cancer death in the world. It is associated with the stimulation of microenvironment, aberrant epigenetic modification, and chronic inflammation. However, few researches discuss the GC molecular progression mechanisms from the perspective of the system level. In this study, we proposed a systems medicine design procedure to identify essential biomarkers and find corresponding drugs for GC. At first, we did big database mining to construct candidate protein-protein interaction network (PPIN) and candidate gene regulation network (GRN). Second, by leveraging the next-generation sequencing (NGS) data, we performed system modeling and applied system identification and model selection to obtain real genome-wide genetic and epigenetic networks (GWGENs). To make the real GWGENs easy to analyze, the principal network projection method was used to extract the core signaling pathways denoted by KEGG pathways. Subsequently, based on the identified biomarkers, we trained a deep neural network of drug-target interaction (DeepDTI) with supervised learning and filtered our candidate drugs considering drug regulation ability and drug sensitivity. With the proposed systematic strategy, we not only shed the light on the progression of GC but also suggested potential multiple-molecule drugs efficiently.

Investigating Pathogenetic Mechanisms of Alzheimer's Disease by Systems Biology Approaches for Drug Discovery.

Alzheimer's disease (AD) is the most common cause of dementia, characterized by progressive cognitive decline and neurodegenerative disorder. Abnormal aggregations of intracellular neurofibrillary tangles (NFTs) and unusual accumulations of extracellular amyloid-ß (Aß) peptides are two important pathological features in AD brains. However, in spite of large-scale clinical studies and computational simulations, the molecular mechanisms of AD development and progression are still unclear. In this study, we divided all of the samples into two groups: early stage (Braak score I-III) and later stage (Braak score IV-VI). By big database mining, the candidate genetic and epigenetic networks (GEN) have been constructed. In order to find out the real GENs for two stages of AD, we performed systems identification and system order detection scheme to prune false positives with the help of corresponding microarray data. Applying the principal network projection (PNP) method, core GENs were extracted from real GENs based on the projection values. By the annotation of KEGG pathway, we could obtain core pathways from core GENs and investigate pathogenetic mechanisms for the early and later stage of AD, respectively. Consequently, according to pathogenetic mechanisms, several potential biomarkers are identified as drug targets for multiple-molecule drug design in the treatment of AD.

Multiple-Molecule Drug Design Based on Systems Biology Approaches and Deep Neural Network to Mitigate Human Skin Aging.

Human skin aging is affected by various biological signaling pathways, microenvironment factors and epigenetic regulations. With the increasing demand for cosmetics and pharmaceuticals to prevent or reverse skin aging year by year, designing multiple-molecule drugs for mitigating skin aging is indispensable. In this study, we developed strategies for systems medicine design based on systems biology methods and deep neural networks. We constructed the candidate genomewide genetic and epigenetic network (GWGEN) via big database mining. After doing systems modeling and applying system identification, system order detection and principle network projection methods with real time-profile microarray data, we could obtain core signaling pathways and identify essential biomarkers based on the skin aging molecular progression mechanisms. Afterwards, we trained a deep neural network of drug-target interaction in advance and applied it to predict the potential candidate drugs based on our identified biomarkers. To narrow down the candidate drugs, we designed two filters considering drug regulation ability and drug sensitivity. With the proposed systems medicine design procedure, we not only shed the light on the skin aging molecular progression mechanisms but also suggested two multiple-molecule drugs for mitigating human skin aging from young adulthood to middle age and middle age to old age, respectively.

Systems Medicine Design for Triple-Negative Breast Cancer and Non-Triple-Negative Breast Cancer Based on Systems Identification and Carcinogenic Mechanisms.

Triple-negative breast cancer (TNBC) is a heterogeneous subtype of breast cancers with poor prognosis. The etiology of triple-negative breast cancer (TNBC) is involved in various biological signal cascades and multifactorial aberrations of genetic, epigenetic and microenvironment. New therapeutic for TNBC is urgently needed because surgery and chemotherapy are the only available modalities nowadays. A better understanding of the molecular mechanisms would be a great challenge because they are triggered by cascade signaling pathways, genetic and epigenetic regulations, and drug-target interactions. This would allow the design of multi-molecule drugs for the TNBC and non-TNBC. In this study, in terms of systems biology approaches, we proposed a systematic procedure for systems medicine design toward TNBC and non-TNBC. For systems biology approaches, we constructed a candidate genome-wide genetic and epigenetic network (GWGEN) by big databases mining and identified real GWGENs of TNBC and non-TNBC assisting with corresponding microarray data by system identification and model order selection methods. After that, we applied the principal network projection (PNP) approach to obtain the core signaling pathways denoted by KEGG pathway of TNBC and non-TNBC. Comparing core signaling pathways of TNBC and non-TNBC, essential carcinogenic biomarkers resulting in multiple cellular dysfunctions including cell proliferation, autophagy, immune response, apoptosis, metastasis, angiogenesis, epithelial-mesenchymal transition (EMT), and cell differentiation could be found. In order to propose potential candidate drugs for the selected biomarkers, we designed filters considering toxicity and regulation ability. With the proposed systematic procedure, we not only shed a light on the differences between carcinogenetic molecular mechanisms of TNBC and non-TNBC but also efficiently proposed candidate multi-molecule drugs including resveratrol, sirolimus, and prednisolone for TNBC and resveratrol, sirolimus, carbamazepine, and verapamil for non-TNBC.

Sex differences in viral entry protein expression and host transcript responses to SARS-CoV-2.

Epidemiological studies suggest that men exhibit a higher mortality rate to COVID-19 than women, yet the underlying biology is largely unknown. Here, we seek to delineate sex differences in the gene expression of viral entry proteins ACE2 and TMPRSS2, and host transcriptional responses to SARS-CoV-2 through large-scale analysis of genomic and clinical data. We first compiled 220,000 human gene expression profiles from three databases and completed the meta-information through machine learning and manual annotation. Large scale analysis of these profiles indicated that male samples show higher expression levels of ACE2 and TMPRSS2 than female samples, especially in the older group (>60 years) and in the kidney. Subsequent analysis of 6,031 COVID-19 patients at Mount Sinai Health System revealed that men have significantly higher creatinine levels, an indicator of impaired kidney function. Further analysis of 782 COVID-19 patient gene expression profiles taken from upper airway and blood suggested men and women present distinct expression changes. Computational deconvolution analysis of these profiles revealed male COVID-19 patients have enriched kidney-specific mesangial cells in blood compared to healthy patients. Together, this study suggests biological differences in the kidney between sexes may contribute to sex disparity in COVID-19.

Investigation of the Genome-Wide Genetic and Epigenetic Networks for Drug Discovery Based on Systems Biology Approaches in Colorectal Cancer.

Colorectal cancer (CRC) is the third most commonly diagnosed type of cancer worldwide. The mechanisms leading to the progression of CRC are involved in both genetic and epigenetic regulations. In this study, we applied systems biology methods to identify potential biomarkers and conduct drug discovery in a computational approach. Using big database mining, we constructed a candidate protein-protein interaction network and a candidate gene regulatory network, combining them into a genome-wide genetic and epigenetic network (GWGEN). With the assistance of system identification and model selection approaches, we obtain real GWGENs for early-stage, mid-stage, and late-stage CRC. Subsequently, we extracted core GWGENs for each stage of CRC from their real GWGENs through a principal network projection method, and projected them to the Kyoto Encyclopedia of Genes and Genomes pathways for further analysis. Finally, we compared these core pathways resulting in different molecular mechanisms in each stage of CRC and identified carcinogenic biomarkers for the design of multiple-molecule drugs to prevent the progression of CRC. Based on the identified gene expression signatures, we suggested potential compounds combined with known CRC drugs to prevent the progression of CRC with querying Connectivity Map (CMap).

Investigation mechanisms between normal, developing and regenerating livers for regenerative liver drug design.

Aim: A systematic multimolecule drug design procedure is proposed for promoting hepatogenesis and liver regeneration. Materials & methods: Genome-wide microarray data including three hepatic conditions are obtained from the GEO database (GSE15238). System modeling and big data mining methods are used to construct real genome-wide genetic-and-epigenetic networks (GWGENs). Then, we extracted the core GWGENs by applying principal network projection on real GWGENs of normal, developing and regenerating livers, respectively. After that, we investigated the significant signal pathways and epigenetic modifications in the core GWGENs to identify potential biomarkers as drug targets. Result & conclusion: A multimolecule drug consisting of sulmazole, clofibrate, colchicine, furazolidone, nadolol, eticlopride and felbinac is proposed to target on novel biomarkers for promoting hepatogenesis and liver regeneration.

Comparing progression molecular mechanisms between lung adenocarcinoma and lung squamous cell carcinoma based on genetic and epigenetic networks: big data mining and genome-wide systems identification.

Non-small-cell lung cancer (NSCLC) is the predominant type of lung cancer in the world. Lung adenocarcinoma (LADC) and lung squamous cell carcinoma (LSCC) are subtypes of NSCLC. We usually regard them as different disease due to their unique molecular characteristics, distinct cells of origin and dissimilar clinical response. However, the differences of genetic and epigenetic progression mechanism between LADC and LSCC are complicated to analyze. Therefore, we applied systems biology approaches and big databases mining to construct genetic and epigenetic networks (GENs) with next-generation sequencing data of LADC and LSCC. In order to obtain the real GENs, system identification and system order detection are conducted on gene regulatory networks (GRNs) and protein-protein interaction networks (PPINs) for each stage of LADC and LSCC. The core GENs were extracted via principal network projection (PNP). Based on the ranking of projection values, we got the core pathways in respect of KEGG pathway. Compared with the core pathways, we found significant differences between microenvironments, dysregulations of miRNAs, epigenetic modifications on certain signaling transduction proteins and target genes in each stage of LADC and LSCC. Finally, we proposed six genetic and epigenetic multiple-molecule drugs to target essential biomarkers in each progression stage of LADC and LSCC, respectively.

Systems Biology Approaches to Investigate Genetic and Epigenetic Molecular Progression Mechanisms for Identifying Gene Expression Signatures in Papillary Thyroid Cancer.

Thyroid cancer is the most common endocrine cancer. Particularly, papillary thyroid cancer (PTC) accounts for the highest proportion of thyroid cancer. Up to now, there are few researches discussing the pathogenesis and progression mechanisms of PTC from the viewpoint of systems biology approaches. In this study, first we constructed the candidate genetic and epigenetic network (GEN) consisting of candidate protein-protein interaction network (PPIN) and candidate gene regulatory network (GRN) by big database mining. Secondly, system identification and system order detection methods were applied to prune candidate GEN via next-generation sequencing (NGS) and DNA methylation profiles to obtain the real GEN. After that, we extracted core GENs from real GENs by the principal network projection (PNP) method. To investigate the pathogenic and progression mechanisms in each stage of PTC, core GEN was denoted in respect of KEGG pathways. Finally, by comparing two successive core signaling pathways of PTC, we not only shed light on the causes of PTC progression, but also identified essential biomarkers with specific gene expression signature. Moreover, based on the identified gene expression signature, we suggested potential candidate drugs to prevent the progression of PTC with querying Connectivity Map (CMap).

Investigating Common Pathogenic Mechanisms between Homo sapiens and Different Strains of Candida albicans for Drug Design: Systems Biology Approach via Two-Sided NGS Data Identification.

Candida albicans (C. albicans) is the most prevalent fungal species. Although it is a healthy microbiota, genetic and epigenetic alterations in host and pathogen, and microenvironment changes would lead to thrush, vaginal yeast infection, and even hematogenously disseminated infection. Despite the fact that cytotoxicity is well-characterized, few studies discuss the genome-wide genetic and epigenetic molecular mechanisms between host and C. albicans. The aim of this study is to identify drug targets and design a multiple-molecule drug to prevent the infection from C. albicans. To investigate the common and specific pathogenic mechanisms in human oral epithelial OKF6/TERT-2 cells during the C. albicans infection in different strains, systems modeling and big databases mining were used to construct candidate host⁻pathogen genetic and epigenetic interspecies network (GEIN). System identification and system order detection are applied on two-sided next generation sequencing (NGS) data to build real host⁻pathogen cross-talk GEINs. Core host⁻pathogen cross-talk networks (HPCNs) are extracted by principal network projection (PNP) method. By comparing with core HPCNs in different strains of C. albicans, common pathogenic mechanisms were investigated and several drug targets were suggested as follows: orf19.5034 (YBP1) with the ability of anti-ROS; orf19.939 (NAM7), orf19.2087 (SAS2), orf19.1093 (FLO8) and orf19.1854 (HHF22) with high correlation to the hyphae growth and pathogen protein interaction; orf19.5585 (SAP5), orf19.5542 (SAP6) and orf19.4519 (SUV3) with the cause of biofilm formation. Eventually, five corresponding compounds-Tunicamycin, Terbinafine, Cerulenin, Tetracycline and Tetrandrine-with three known drugs could be considered as a potential multiple-molecule drug for therapeutic treatment of C. albicans.

Effect of backrest height on wheelchair propulsion biomechanics for level and uphill conditions.

OBJECTIVE: To evaluate the effect of backrest height on wheelchair propulsion kinematics and kinetics. DESIGN: An intervention study with repeated measures. SETTING: University laboratory. PARTICIPANTS: Convenience sample included manual wheelchair users (N=36; 26 men and 10 women) with spinal cord injuries ranging from T8 to L2. INTERVENTION: Participants propelled on a motor-driven treadmill for 2 conditions (level and slope of 3°) at a constant speed of 0.9 m/s while using in turn a sling backrest fixed at 40.6 cm (16 in) high (high backrest) and a lower height set at 50% trunk length (low backrest). MAIN OUTCOME MEASURES: Cadence, stroke angle, peak shoulder extension angle, shoulder flexion/extension range of motion, and mechanical effective force. RESULTS: Pushing with the low backrest height enabled greater range of shoulder motion (P<.01), increased stroke angle (P<.01), push time (P<.01), and reduced cadence (P=.01) regardless of whether the treadmill was level or sloped. CONCLUSIONS: A lower cadence can be achieved when pushing with a lower backrest, which decreases the risk of developing upper-limb overuse related injuries. However, postural support, comfort, and other activities of daily living must also be considered when selecting a backrest height for active, long-term wheelchair users. The improvements found when using the low backrest were found regardless of slope type. Pushing uphill demanded significantly higher resultant and tangential force, torque, mechanical effective force, and cadence.