Discriminating the single-cell gene regulatory networks of human pancreatic islets: A novel deep learning application

A novel interpretable deep transfer learning combining diverse learnable parameters for improved T2D prediction based on single-cell gene regulatory networks

Accurate deep learning (DL) models to predict type 2 diabetes (T2D) are concerned not only with targeting the discrimination task but also with learning useful feature representation. However, existing DL tools are far from perfect and do not provide appropriate interpretation as a guideline to explain and promote superior performance in the target task. Therefore, we provide an interpretable approach for our presented deep transfer learning (DTL) models to overcome such drawbacks, working as follows. We utilize several pre-trained models including SEResNet152, and SEResNeXT101. Then, we transfer knowledge from pre-trained models via keeping the weights in the convolutional base (i.e., feature extraction part) while modifying the classification part with the use of Adam optimizer to deal with classifying healthy controls and T2D based on single-cell gene regulatory network (SCGRN) images. Another DTL models work in a similar manner but just with keeping weights of the bottom layers in the feature extraction unaltered while updating weights of consecutive layers through training from scratch. Experimental results on the whole 224 SCGRN images using five-fold cross-validation show that our model (TFeSEResNeXT101) achieving the highest average balanced accuracy (BAC) of 0.97 and thereby significantly outperforming the baseline that resulted in an average BAC of 0.86. Moreover, the simulation study demonstrated that the superiority is attributed to the distributional conformance of model weight parameters obtained with Adam optimizer when coupled with weights from a pre-trained model.

Recent Developments in Islet Biology: A Review With Patient Perspectives

Navigating the coronavirus disease-2019 (COVID-19, now COVID) pandemic has required resilience and creativity worldwide. Despite early challenges to productivity, more than 2,000 peer-reviewed articles on islet biology were published in 2021. Herein, we highlight noteworthy advances in islet research between January 2021 and April 2022, focussing on 5 areas. First, we discuss new insights into the role of glucokinase, mitogen-activated protein kinase-kinase/extracellular signal-regulated kinase and mitochondrial function on insulin secretion from the pancreatic β cell, provided by new genetically modified mouse models and live imaging. We then discuss a new connection between lipid handling and improved insulin secretion in the context of glucotoxicity, focussing on fatty acid-binding protein 4 and fetuin-A. Advances in high-throughput "omic" analysis evolved to where one can generate more finely tuned genetic and molecular profiles within broad classifications of type 1 diabetes and type 2 diabetes. Next, we highlight breakthroughs in diabetes treatment using stem cell-derived β cells and innovative strategies to improve islet survival posttransplantation. Last, we update our understanding of the impact of severe acute respiratory syndrome-coronavirus-2 infection on pancreatic islet function and discuss current evidence regarding proposed links between COVID and new-onset diabetes. We address these breakthroughs in 2 settings: one for a scientific audience and the other for the public, particularly those living with or affected by diabetes. Bridging biomedical research in diabetes to the community living with or affected by diabetes, our partners living with type 1 diabetes or type 2 diabetes also provide their perspectives on these latest advances in islet biology.

Artificial Intelligence for the Prediction and Early Diagnosis of Pancreatic Cancer: Scoping Review (Preprint).. [DOI: 10.2196/preprints.44248] [Citation(s) in RCA: 0] [Impact Index Per Article: 0]

BACKGROUND

Pancreatic cancer is the 12th most common cancer worldwide, with an overall survival rate of 4.9%. Early diagnosis of pancreatic cancer is essential for timely treatment and survival. Artificial intelligence (AI) provides advanced models and algorithms for better diagnosis of pancreatic cancer.

OBJECTIVE

This study aims to explore AI models used for the prediction and early diagnosis of pancreatic cancers as reported in the literature.

METHODS

A scoping review was conducted and reported in line with the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines. PubMed, Google Scholar, Science Direct, BioRXiv, and MedRxiv were explored to identify relevant articles. Study selection and data extraction were independently conducted by 2 reviewers. Data extracted from the included studies were synthesized narratively.

RESULTS

Of the 1185 publications, 30 studies were included in the scoping review. The included articles reported the use of AI for 6 different purposes. Of these included articles, AI techniques were mostly used for the diagnosis of pancreatic cancer (14/30, 47%). Radiological images (14/30, 47%) were the most frequently used data in the included articles. Most of the included articles used data sets with a size of <1000 samples (11/30, 37%). Deep learning models were the most prominent branch of AI used for pancreatic cancer diagnosis in the studies, and the convolutional neural network was the most used algorithm (18/30, 60%). Six validation approaches were used in the included studies, of which the most frequently used approaches were k-fold cross-validation (10/30, 33%) and external validation (10/30, 33%). A higher level of accuracy (99%) was found in studies that used support vector machine, decision trees, and k-means clustering algorithms.

CONCLUSIONS

This review presents an overview of studies based on AI models and algorithms used to predict and diagnose pancreatic cancer patients. AI is expected to play a vital role in advancing pancreatic cancer prediction and diagnosis. Further research is required to provide data that support clinical decisions in health care.

PredMHC: An Effective Predictor of Major Histocompatibility Complex Using Mixed Features

The major histocompatibility complex (MHC) is a large locus on vertebrate DNA that contains a tightly linked set of polymorphic genes encoding cell surface proteins essential for the adaptive immune system. The groups of proteins encoded in the MHC play an important role in the adaptive immune system. Therefore, the accurate identification of the MHC is necessary to understand its role in the adaptive immune system. An effective predictor called PredMHC is established in this study to identify the MHC from protein sequences. Firstly, PredMHC encoded a protein sequence with mixed features including 188D, APAAC, KSCTriad, CKSAAGP, and PAAC. Secondly, three classifiers including SGD, SMO, and random forest were trained on the mixed features of the protein sequence. Finally, the prediction result was obtained by the voting of the three classifiers. The experimental results of the 10-fold cross-validation test in the training dataset showed that PredMHC can obtain 91.69% accuracy. Experimental results on comparison with other features, classifiers, and existing methods showed the effectiveness of PredMHC in predicting the MHC.

CCDC134 as a Prognostic-Related Biomarker in Breast Cancer Correlating With Immune Infiltrates

Background

The expression of Coiled-Coil Domain Containing 134(CCDC134) is up-regulated in different pan-cancer species. However, its prognostic value and correlation with immune infiltration in breast cancer are unclear. Therefore, we evaluated the prognostic role of CCDC134 in breast cancer and its correlation with immune invasion.

Methods

We downloaded the transcription profile of CCDC134 between breast cancer and normal tissues from the Cancer Genome Atlas (TCGA). CCDC134 protein expression was assessed by the Clinical Proteomic Cancer Analysis Consortium (CPTAC) and the Human Protein Atlas. Gene set enrichment analysis (GSEA) was also used for pathway analysis. Receiver operating characteristic (ROC) curve was used to differentiate breast cancer from adjacent normal tissues. Kaplan-Meier method was used to evaluate the effect of CCDC134 on survival rate. The protein-protein interaction (PPI) network is built from STRING. Function expansion analysis is performed using the ClusterProfiler package. Through tumor Immune Estimation Resource (TIMER) and tumor Immune System Interaction database (TISIDB) to determine the relationship between CCDC134 expression level and immune infiltration. CTD database is used to predict drugs that inhibit CCDC134 and PubChem database is used to determine the molecular structure of identified drugs.

Results

The expression of CCDC134 in breast cancer tissues was significantly higher than that of CCDC134 mRNA expression in adjacent normal tissues. ROC curve analysis showed that the AUC value of CCDC134 was 0.663. Kaplan-meier survival analysis showed that patients with high CCDC134 had a lower prognosis (57.27 months vs 36.96 months, P = 2.0E-6). Correlation analysis showed that CCDC134 mRNA expression was associated with tumor purity immune invasion. In addition, CTD database analysis identified abrine, Benzo (A) Pyrene, bisphenol A, Soman, Sunitinib, Tetrachloroethylene, Valproic Acid as seven targeted therapy drugs that may be effective treatments for seven targeted therapeutics. It may be an effective treatment for inhibiting CCDC134.

Conclusion

In breast cancer, upregulated CCDC134 is significantly associated with lower survival and immune infiltrates invasion. Our study suggests that CCDC134 can serve as a biomarker of poor prognosis and a potential immunotherapy target in breast cancer. Seven drugs with significant potential to inhibit CCDC134 were identified.

Inference on the structure of gene regulatory networks

In this paper, we conduct theoretical analyses on inferring the structure of gene regulatory networks. Depending on the experimental method and data type, the inference problem is classified into 20 different scenarios. For each scenario, we discuss the problem that with enough data, under what assumptions, what can be inferred about the structure. For scenarios that have been covered in the literature, we provide a brief review. For scenarios that have not been covered in literature, if the structure can be inferred, we propose new mathematical inference methods and evaluate them on simulated data. Otherwise, we prove that the structure cannot be inferred.