Classification of benign and malignant subtypes of breast cancer histopathology imaging using hybrid CNN-LSTM based transfer learning.

BACKGROUND: Grading of cancer histopathology slides requires more pathologists and expert clinicians as well as it is time consuming to look manually into whole-slide images. Hence, an automated classification of histopathological breast cancer sub-type is useful for clinical diagnosis and therapeutic responses. Recent deep learning methods for medical image analysis suggest the utility of automated radiologic imaging classification for relating disease characteristics or diagnosis and patient stratification. METHODS: To develop a hybrid model using the convolutional neural network (CNN) and the long short-term memory recurrent neural network (LSTM RNN) to classify four benign and four malignant breast cancer subtypes. The proposed CNN-LSTM leveraging on ImageNet uses a transfer learning approach in classifying and predicting four subtypes of each. The proposed model was evaluated on the BreakHis dataset comprises 2480 benign and 5429 malignant cancer images acquired at magnifications of 40×, 100×, 200× and 400×. RESULTS: The proposed hybrid CNN-LSTM model was compared with the existing CNN models used for breast histopathological image classification such as VGG-16, ResNet50, and Inception models. All the models were built using three different optimizers such as adaptive moment estimator (Adam), root mean square propagation (RMSProp), and stochastic gradient descent (SGD) optimizers by varying numbers of epochs. From the results, we noticed that the Adam optimizer was the best optimizer with maximum accuracy and minimum model loss for both the training and validation sets. The proposed hybrid CNN-LSTM model showed the highest overall accuracy of 99% for binary classification of benign and malignant cancer, and, whereas, 92.5% for multi-class classifier of benign and malignant cancer subtypes, respectively. CONCLUSION: To conclude, the proposed transfer learning approach outperformed the state-of-the-art machine and deep learning models in classifying benign and malignant cancer subtypes. The proposed method is feasible in classification of other cancers as well as diseases.

Identification of Novel Ribonucleotide Reductase Inhibitors for Therapeutic Application in Bile Tract Cancer: An Advanced Pharmacoinformatics Study.

Biliary tract cancer (BTC) is constituted by a heterogeneous group of malignant tumors that may develop in the biliary tract, and it is the second most common liver cancer. Human ribonucleotide reductase M1 (hRRM1) has already been proven to be a potential BTC target. In the current study, a de novo design approach was used to generate novel and effective chemical therapeutics for BTC. A set of comprehensive pharmacoinformatics approaches was implemented and, finally, seventeen potential molecules were found to be effective for the modulation of hRRM1 activity. Molecular docking, negative image-based ShaEP scoring, absolute binding free energy, in silico pharmacokinetics, and toxicity assessments corroborated the potentiality of the selected molecules. Almost all molecules showed higher affinity in comparison to gemcitabine and naphthyl salicylic acyl hydrazone (NSAH). On binding interaction analysis, a number of critical amino acids was found to hold the molecules at the active site cavity. The molecular dynamics (MD) simulation study also indicated the stability between protein and ligands. High negative MM-GBSA (molecular mechanics generalized Born and surface area) binding free energy indicated the potentiality of the molecules. Therefore, the proposed molecules might have the potential to be effective therapeutics for the management of BTC.

Identification of Potential Cytochrome P450 3A5 Inhibitors: An Extensive Virtual Screening through Molecular Docking, Negative Image-Based Screening, Machine Learning and Molecular Dynamics Simulation Studies.

Cytochrome P450 3A5 (CYP3A5) is one of the crucial CYP family members and has already proven to be an important drug target for cardiovascular diseases. In the current study, the PubChem database was screened through molecular docking and high-affinity molecules were adopted for further assessment. A negative image-based (NIB) model was used for a similarity search by considering the complementary shape and electrostatics of the target and small molecules. Further, the molecules were segregated into active and inactive groups through six machine learning (ML) matrices. The active molecules found in each ML model were used for in silico pharmacokinetics and toxicity assessments. A total of five molecules followed the acceptable pharmacokinetics and toxicity profiles. Several potential binding interactions between the proposed molecules and CYP3A5 were observed. The dynamic behavior of the selected molecules in the CYP3A5 was explored through a molecular dynamics (MD) simulation study. Several parameters obtained from the MD simulation trajectory explained the stability of the protein-ligand complexes in dynamic states. The high binding affinity of each molecule was revealed by the binding free energy calculation through the MM-GBSA methods. Therefore, it can be concluded that the proposed molecules might be potential CYP3A5 molecules for therapeutic application in cardiovascular diseases subjected to in vitro/in vivo validations.

Screening of ß1- and ß2-Adrenergic Receptor Modulators through Advanced Pharmacoinformatics and Machine Learning Approaches.

Cardiovascular diseases (CDs) are a major concern in the human race and one of the leading causes of death worldwide. ß-Adrenergic receptors (ß1-AR and ß2-AR) play a crucial role in the overall regulation of cardiac function. In the present study, structure-based virtual screening, machine learning (ML), and a ligand-based similarity search were conducted for the PubChem database against both ß1- and ß2-AR. Initially, all docked molecules were screened using the threshold binding energy value. Molecules with a better binding affinity were further used for segregation as active and inactive through ML. The pharmacokinetic assessment was carried out on molecules retained in the above step. Further, similarity searching of the ChEMBL and DrugBank databases was performed. From detailed analysis of the above data, four compounds for each of ß1- and ß2-AR were found to be promising in nature. A number of critical ligand-binding amino acids formed potential hydrogen bonds and hydrophobic interactions. Finally, a molecular dynamics (MD) simulation study of each molecule bound with the respective target was performed. A number of parameters obtained from the MD simulation trajectories were calculated and substantiated the stability between the protein-ligand complex. Hence, it can be postulated that the final molecules might be crucial for CDs subjected to experimental validation.

Variation in human chromosome 21 ribosomal RNA genes characterized by TAR cloning and long-read sequencing.

Despite the key role of the human ribosome in protein biosynthesis, little is known about the extent of sequence variation in ribosomal DNA (rDNA) or its pre-rRNA and rRNA products. We recovered ribosomal DNA segments from a single human chromosome 21 using transformation-associated recombination (TAR) cloning in yeast. Accurate long-read sequencing of 13 isolates covering ∼0.82 Mb of the chromosome 21 rDNA complement revealed substantial variation among tandem repeat rDNA copies, several palindromic structures and potential errors in the previous reference sequence. These clones revealed 101 variant positions in the 45S transcription unit and 235 in the intergenic spacer sequence. Approximately 60% of the 45S variants were confirmed in independent whole-genome or RNA-seq data, with 47 of these further observed in mature 18S/28S rRNA sequences. TAR cloning and long-read sequencing enabled the accurate reconstruction of multiple rDNA units and a new, high-quality 44 838 bp rDNA reference sequence, which we have annotated with variants detected from chromosome 21 of a single individual. The large number of variants observed reveal heterogeneity in human rDNA, opening up the possibility of corresponding variations in ribosome dynamics.

Analysis of Circular RNAs Using the Web Tool CircInteractome.

Circular RNAs (circRNAs) are generated through nonlinear back splicing, during which the 5' and 3' ends are covalently joined. Consequently, the lack of free ends makes them very stable compared to their counterpart linear RNAs. By selectively interacting with microRNAs and RNA-binding proteins (RBPs), circRNAs have been shown to influence gene expression programs. We designed a web tool, CircInteractome, in order to (1) explore potential interactions of circRNAs with RBPs, (2) design specific divergent primers to detect circRNAs, (3) study tissue- and cell-specific circRNAs, (4) identify gene-specific circRNAs, (5) explore potential miRNAs interacting with circRNAs, and (6) design specific siRNAs to silence circRNAs. Here, we review the CircInteractome tool and explain recent updates to the site. The database is freely accessible at http://circinteractome.nia.nih.gov .

WIG1 is crucial for AGO2-mediated ACOT7 mRNA silencing via miRNA-dependent and -independent mechanisms.

RNA-binding proteins (RBPs) are involved in mRNA splicing, maturation, transport, translation, storage and turnover. Here, we identified ACOT7 mRNA as a novel target of human WIG1. ACOT7 mRNA decay was triggered by the microRNA miR-9 in a WIG1-dependent manner via classic recruitment of Argonaute 2 (AGO2). Interestingly, AGO2 was also recruited to ACOT7 mRNA in a WIG1-dependent manner in the absence of miR-9, which indicates an alternative model whereby WIG1 controls AGO2-mediated gene silencing. The WIG1-AGO2 complex attenuated translation initiation via an interaction with translation initiation factor 5B (eIF5B). These results were confirmed using a WIG1 tethering system based on the MS2 bacteriophage coat protein and a reporter construct containing an MS2-binding site, and by immunoprecipitation of WIG1 and detection of WIG1-associated proteins using liquid chromatography-tandem mass spectrometry. We also identified WIG1-binding motifs using photoactivatable ribonucleoside-enhanced crosslinking and immunoprecipitation analyses. Altogether, our data indicate that WIG1 governs the miRNA-dependent and the miRNA-independent recruitment of AGO2 to lower the stability of and suppress the translation of ACOT7 mRNA.

High-purity circular RNA isolation method (RPAD) reveals vast collection of intronic circRNAs.

High-throughput RNA sequencing methods coupled with specialized bioinformatic analyses have recently uncovered tens of thousands of unique circular (circ)RNAs, but their complete sequences, genes of origin and functions are largely unknown. Given that circRNAs lack free ends and are thus relatively stable, their association with microRNAs (miRNAs) and RNA-binding proteins (RBPs) can influence gene expression programs. While exoribonuclease treatment is widely used to degrade linear RNAs and enrich circRNAs in RNA samples, it does not efficiently eliminate all linear RNAs. Here, we describe a novel method for the isolation of highly pure circRNA populations involving RNase R treatment followed by Polyadenylation and poly(A)+ RNA Depletion (RPAD), which removes linear RNA to near completion. High-throughput sequencing of RNA prepared using RPAD from human cervical carcinoma HeLa cells and mouse C2C12 myoblasts led to two surprising discoveries: (i) many exonic circRNA (EcircRNA) isoforms share an identical backsplice sequence but have different body sizes and sequences, and (ii) thousands of novel intronic circular RNAs (IcircRNAs) are expressed in cells. In sum, isolating high-purity circRNAs using the RPAD method can enable quantitative and qualitative analyses of circRNA types and sequence composition, paving the way for the elucidation of circRNA functions.

Identification of HuR target circular RNAs uncovers suppression of PABPN1 translation by CircPABPN1.

HuR influences gene expression programs and hence cellular phenotypes by binding to hundreds of coding and noncoding linear RNAs. However, whether HuR binds to circular RNAs (circRNAs) and impacts on their function is unknown. Here, we have identified en masse circRNAs binding HuR in human cervical carcinoma HeLa cells. One of the most prominent HuR target circRNAs was hsa_circ_0031288, renamed CircPABPN1 as it arises from the PABPN1 pre-mRNA. Further analysis revealed that HuR did not influence CircPABPN1 abundance; interestingly, however, high levels of CircPABPN1 suppressed HuR binding to PABPN1 mRNA. Evaluation of PABPN1 mRNA polysomes indicated that PABPN1 translation was modulated positively by HuR and hence negatively by CircPABPN1. We propose that the extensive binding of CircPABPN1 to HuR prevents HuR binding to PABPN1 mRNA and lowers PABPN1 translation, providing the first example of competition between a circRNA and its cognate mRNA for an RBP that affects translation.

Emergence of undifferentiated colonies from mouse embryonic stem cells undergoing differentiation by retinoic acid treatment.

Retinoic acid (RA) is one of the most potent inducers of differentiation of mouse embryonic stem cells (ESCs). However, previous studies show that RA treatment of cells cultured in the presence of a leukemia inhibitory factor (LIF) also result in the upregulation of a gene called Zscan4, whose transient expression is a marker for undifferentiated ESCs. We explored the balance between these two seemingly antagonistic effects of RA. ESCs indeed differentiated in the presence of LIF after RA treatment, but colonies of undifferentiated ESCs eventually emerged from these differentiated cells - even in the presence of RA. These colonies, named secondary colonies, consist of three cell types: typical undifferentiated ESCs expressing pluripotency genes such as Pou5f1, Sox2, and Nanog; cells expressing Zscan4; and endodermal-like cells located at the periphery of the colony. The capacity to form secondary colonies was confirmed for all eight tested ESC lines. Cells from the secondary colonies - after transfer to the standard ESC medium - retained pluripotency, judged by their strong alkaline phosphatase (ALP) staining, typical colony morphology, gene expression profile, stable karyotype, capacity to differentiate into all three germ layers in embryoid body formation assays, and successful contribution to chimeras after injection into blastocysts. Based on flow cytometry analysis (FACS), the proportion of Zscan4-positive cells in secondary colonies was higher than in standard ESC colonies, which may explain the capacity of ESCs to resist the differentiating effects of RA and instead form secondary colonies of undifferentiated ESCs. This hypothesis is supported by cell-lineage tracing analysis, which showed that most cells in the secondary colonies were descendents of cells transiently expressing Zscan4.

HuR and GRSF1 modulate the nuclear export and mitochondrial localization of the lncRNA RMRP.

Some mitochondrial long noncoding RNAs (lncRNAs) are encoded by nuclear DNA, but the mechanisms that mediate their transport to mitochondria are poorly characterized. Using affinity RNA pull-down followed by mass spectrometry analysis, we found two RNA-binding proteins (RBPs), HuR (human antigen R) and GRSF1 (G-rich RNA sequence-binding factor 1), that associated with the nuclear DNA-encoded lncRNA RMRP and mobilized it to mitochondria. In cultured human cells, HuR bound RMRP in the nucleus and mediated its CRM1 (chromosome region maintenance 1)-dependent export to the cytosol. After RMRP was imported into mitochondria, GRSF1 bound RMRP and increased its abundance in the matrix. Loss of GRSF1 lowered the mitochondrial levels of RMRP, in turn suppressing oxygen consumption rates and modestly reducing mitochondrial DNA replication priming. Our findings indicate that RBPs HuR and GRSF1 govern the cytoplasmic and mitochondrial localization of the lncRNA RMRP, which is encoded by nuclear DNA but has key functions in mitochondria.

Generation and gene expression profiling of 48 transcription-factor-inducible mouse embryonic stem cell lines.

Mouse embryonic stem cells (ESCs) can differentiate into a wide range - and possibly all cell types in vitro, and thus provide an ideal platform to study systematically the action of transcription factors (TFs) in cell differentiation. Previously, we have generated and analyzed 137 TF-inducible mouse ESC lines. As an extension of this "NIA Mouse ESC Bank," we generated and characterized 48 additional mouse ESC lines, in which single TFs in each line could be induced in a doxycycline-controllable manner. Together, with the previous ESC lines, the bank now comprises 185 TF-manipulable ESC lines (>10% of all mouse TFs). Global gene expression (transcriptome) profiling revealed that the induction of individual TFs in mouse ESCs for 48 hours shifts their transcriptomes toward specific differentiation fates (e.g., neural lineages by Myt1 Isl1, and St18; mesodermal lineages by Pitx1, Pitx2, Barhl2, and Lmx1a; white blood cells by Myb, Etv2, and Tbx6, and ovary by Pitx1, Pitx2, and Dmrtc2). These data also provide and lists of inferred target genes of each TF and possible functions of these TFs. The results demonstrate the utility of mouse ESC lines and their transcriptome data for understanding the mechanism of cell differentiation and the function of TFs.

Novel RNA-binding activity of MYF5 enhances Ccnd1/Cyclin D1 mRNA translation during myogenesis.

Skeletal muscle contains long multinucleated and contractile structures known as muscle fibers, which arise from the fusion of myoblasts into multinucleated myotubes during myogenesis. The myogenic regulatory factor (MRF) MYF5 is the earliest to be expressed during myogenesis and functions as a transcription factor in muscle progenitor cells (satellite cells) and myocytes. In mouse C2C12 myocytes, MYF5 is implicated in the initial steps of myoblast differentiation into myotubes. Here, using ribonucleoprotein immunoprecipitation (RIP) analysis, we discovered a novel function for MYF5 as an RNA-binding protein which associated with a subset of myoblast mRNAs. One prominent MYF5 target was Ccnd1 mRNA, which encodes the key cell cycle regulator CCND1 (Cyclin D1). Biotin-RNA pulldown, UV-crosslinking and gel shift experiments indicated that MYF5 was capable of binding the 3' untranslated region (UTR) and the coding region (CR) of Ccnd1 mRNA. Silencing MYF5 expression in proliferating myoblasts revealed that MYF5 promoted CCND1 translation and modestly increased transcription of Ccnd1 mRNA. Accordingly, overexpressing MYF5 in C2C12 cells upregulated CCND1 expression while silencing MYF5 reduced myoblast proliferation as well as differentiation of myoblasts into myotubes. Moreover, MYF5 silencing reduced myogenesis, while ectopically restoring CCND1 abundance partially rescued the decrease in myogenesis seen after MYF5 silencing. We propose that MYF5 enhances early myogenesis in part by coordinately elevating Ccnd1 transcription and Ccnd1 mRNA translation.

CircInteractome: A web tool for exploring circular RNAs and their interacting proteins and microRNAs.

Circular RNAs (circRNAs) are widely expressed in animal cells, but their biogenesis and functions are poorly understood. CircRNAs have been shown to act as sponges for miRNAs and may also potentially sponge RNA-binding proteins (RBPs) and are thus predicted to function as robust posttranscriptional regulators of gene expression. The joint analysis of large-scale transcriptome data coupled with computational analyses represents a powerful approach to elucidate possible biological roles of ribonucleoprotein (RNP) complexes. Here, we present a new web tool, CircInteractome (circRNA interactome), for mapping RBP- and miRNA-binding sites on human circRNAs. CircInteractome searches public circRNA, miRNA, and RBP databases to provide bioinformatic analyses of binding sites on circRNAs and additionally analyzes miRNA and RBP sites on junction and junction-flanking sequences. CircInteractome also allows the user the ability to (1) identify potential circRNAs which can act as RBP sponges, (2) design junction-spanning primers for specific detection of circRNAs of interest, (3) design siRNAs for circRNA silencing, and (4) identify potential internal ribosomal entry sites (IRES). In sum, the web tool CircInteractome, freely accessible at http://circinteractome.nia.nih.gov, facilitates the analysis of circRNAs and circRNP biology.

Transient bursts of Zscan4 expression are accompanied by the rapid derepression of heterochromatin in mouse embryonic stem cells.

Mouse embryonic stem cells (mESCs) have a remarkable capacity to maintain normal genome stability and karyotype in culture. We previously showed that infrequent bursts of Zscan4 expression (Z4 events) are important for the maintenance of telomere length and genome stability in mESCs. However, the molecular details of Z4 events remain unclear. Here we show that Z4 events involve unexpected transcriptional derepression in heterochromatin regions that usually remain silent. During a Z4 event, we see rapid derepression and rerepression of heterochromatin leading to a burst of transcription that coincides with transient histone hyperacetylation and DNA demethylation, clustering of pericentromeric heterochromatin around the nucleolus, and accumulation of activating and repressive chromatin remodelling complexes. This heterochromatin-based transcriptional activity suggests that mESCs may maintain their extraordinary genome stability at least in part by transiently resetting their heterochromatin.

7SL RNA represses p53 translation by competing with HuR.

Noncoding RNAs (ncRNAs) and RNA-binding proteins are potent post-transcriptional regulators of gene expression. The ncRNA 7SL is upregulated in cancer cells, but its impact upon the phenotype of cancer cells is unknown. Here, we present evidence that 7SL forms a partial hybrid with the 3'-untranslated region (UTR) of TP53 mRNA, which encodes the tumor suppressor p53. The interaction of 7SL with TP53 mRNA reduced p53 translation, as determined by analyzing p53 expression levels, nascent p53 translation and TP53 mRNA association with polysomes. Silencing 7SL led to increased binding of HuR to TP53 mRNA, an interaction that led to the promotion of p53 translation and increased p53 abundance. We propose that the competition between 7SL and HuR for binding to TP53 3'UTR contributes to determining the magnitude of p53 translation, in turn affecting p53 levels and the growth-suppressive function of p53. Our findings suggest that targeting 7SL may be effective in the treatment of cancers with reduced p53 levels.

Chromatin properties of regulatory DNA probed by manipulation of transcription factors.

Transcription factors (TFs) bind to DNA and regulate the transcription of nearby genes. However, only a small fraction of TF binding sites have such regulatory effects. Here we search for the predictors of functional binding sites by carrying out a systematic computational screening of a variety of contextual factors (histone modifications, nuclear lamin-bindings, and cofactor bindings). We used regression analysis to test if contextual factors are associated with upregulation or downregulation of neighboring genes following the induction or knockdown of the 9 TFs in mouse embryonic stem (ES) cells. Functional TF binding sites appeared to be either active (i.e., bound by P300, CHD7, mediator, cohesin, and SWI/SNF) or repressed (i.e., with H3K27me3 histone marks and bound by Polycomb factors). Active binding sites mediated the downregulation of nearby genes upon knocking down the activating TFs or inducing repressors. Repressed TF binding sites mediated the upregulation of nearby genes (e.g., poised developmental regulators) upon inducing TFs. In addition, repressed binding sites mediated repressive effects of TFs, identified by the downregulation of target genes after the induction of TFs or by the upregulation of target genes after the knockdown of TFs. The contextual factors associated with functions of DNA-bound TFs were used to improve the identification of candidate target genes regulated by TFs.

Systematic repression of transcription factors reveals limited patterns of gene expression changes in ES cells.

Networks of transcription factors (TFs) are thought to determine and maintain the identity of cells. Here we systematically repressed each of 100 TFs with shRNA and carried out global gene expression profiling in mouse embryonic stem (ES) cells. Unexpectedly, only the repression of a handful of TFs significantly affected transcriptomes, which changed in two directions/trajectories: one trajectory by the repression of either Pou5f1 or Sox2; the other trajectory by the repression of either Esrrb, Sall4, Nanog, or Tcfap4. The data suggest that the trajectories of gene expression change are already preconfigured by the gene regulatory network and roughly correspond to extraembryonic and embryonic fates of cell differentiation, respectively. These data also indicate the robustness of the pluripotency gene network, as the transient repression of most TFs did not alter the transcriptomes.

SHAVE: shrinkage estimator measured for multiple visits increases power in GWAS of quantitative traits.

Measurement error and biological variability generate distortions in quantitative phenotypic data. In longitudinal studies with repeated measurements, the multiple measurements provide a route to reduce noise and correspondingly increase the strength of signals in genome-wide association studies (GWAS).To optimize noise correction, we have developed Shrunken Average (SHAVE), an approach using a Bayesian Shrinkage estimator. This estimator uses regression toward the mean for every individual as a function of (1) their average across visits; (2) their number of visits; and (3) the correlation between visits. Computer simulations support an increase in power, with results very similar to those expected by the assumptions of the model. The method was applied to a real data set for 14 anthropomorphic traits in ∼6000 individuals enrolled in the SardiNIA project, with up to three visits (measurements) for each participant. Results show that additional measurements have a large impact on the strength of GWAS signals, especially when participants have different number of visits, with SHAVE showing a clear increase in power relative to single visits. In addition, we have derived a relation to assess the improvement in power as a function of number of visits and correlation between visits. It can also be applied in the optimization of experimental designs or usage of measuring devices. SHAVE is fast and easy to run, written in R and freely available online.

Generation of mouse ES cell lines engineered for the forced induction of transcription factors.

Here we report the generation and characterization of 84 mouse ES cell lines with doxycycline-controllable transcription factors (TFs) which, together with the previous 53 lines, cover 7-10% of all TFs encoded in the mouse genome. Global gene expression profiles of all 137 lines after the induction of TFs for 48 hrs can associate each TF with the direction of ES cell differentiation, regulatory pathways, and mouse phenotypes. These cell lines and microarray data provide building blocks for a variety of future biomedical research applications as a community resource.