bioRxiv Science⌕ Search

Biology subjects

Singh, K. K.

Publications and source records attributed to Singh, K. K..

At least 19 recordsLinked to original sources

From Plants to Patients: Mitochondrial Stress Signaling as a Systems Framework for Human Disease Vulnerability

Mitochondria integrate bioenergetic metabolism, redox control, genome maintenance, and stress signaling across all eukaryotes. Although plant and human mitochondria diverged substantially during evolution, both systems retain systems-level principles for sensing mitochondrial dysfunction and communicating stress signals to the nucleus. Here, we develop an integrative comparative in silico framework to evaluate whether plant mitochondrial stress signaling can provide a useful conceptual model for interpreting human mitochondrial disease vulnerability. Core Arabidopsis thaliana regulators representing alternative respiration, mitochondrial retrograde signaling, translational stress control, and genome surveillance were compared with functionally analogous human regulators involved in integrated stress response (ISR) signaling, mitochondrial DNA maintenance, and mitochondrial disease phenotypes. Domain architecture, protein-protein interaction topology, enrichment profiles, disease-gene associations, and promoter motif architecture were integrated to assess cross-kingdom convergence at the level of stress-response organization rather than direct orthologs. The plant network formed a compact AOX-NAC-centered stress module associated with respiratory flexibility and retrograde signaling, whereas the human network displayed expanded ISR and mtDNA maintenance modules enriched for mitochondrial disease associations. Promoter motif analyses further indicated lineage-specific transcription factor signatures but broadly comparable stress-responsive regulatory logic. Collectively, these results support the concept that plant mitochondrial stress systems represent simplified resilience-oriented architectures that can help generate experimentally testable hypotheses about failure points in human mitochondrial stress responses.

bioinformatics↗

Fabrication of a Parali (Rice Straws) Biomass & Herbal Based Filtration System for Sustainable Rural Water Purification

Access to safe drinking water remains a serious challenge in rural India, where traditional chemical and membrane-based filtering systems are often unavailable due to their high cost, energy requirements, and maintenance demands. This work develops and validates a low-cost, eco-friendly herbal filtration system for Mubarakpur hamlet by combining the antibacterial properties of plant extracts with natural filter media. The physical, chemical, and microbiological properties of the water samples collected from ponds, wells, sewage, and tap sources included turbidity, pH, COD, BOD, heavy metals, and bacterial load. Herbal extracts derived from Neem (Azadirachta indica), Bael (Aegle marmelos), Lemon, Amla, Moringa, Garlic, and Ginger had bactericidal effects when tested for antimicrobial activity; however, the peels of Bael and Lemon exhibited the best inhibition of Streptobacillus species. A six-layer prototype filtration device including cotton cloth, charcoal, parali biomass, gravel, sand, and a layer of herbal extract was created as a result of these findings. The extracts of lemon and bael exhibited the greatest inhibition of Streptobacillus species when tested for antibacterial activity. A six-layer prototype filtration device including cotton cloth, charcoal, parali biomass, gravel, sand, and a layer of herbal extract was created as a result of these findings. Laboratory testing of the small prototype revealed significant improvements in water quality: the microbial load dropped from 2,000,000 CFU/mL to 0 CFU/mL, the turbidity went from 300 NTU to 0.8 NTU, the COD decreased from 1000 mg/L to 9 mg/L, and the BOD decreased from 200 mg/L to 3.5 mg/L. Lead levels in tap water (0.02 mg/L) were also reduced to undetectable levels.

bioengineering↗

Title - ATG7 vs. ATG5: Distinct Autophagy Pathways Shaping TGF-β Signaling and Endothelial Function

Autophagy and transforming growth factor-beta (TGF-{beta}) signalling are critical cellular processes that maintain homeostasis and regulate various physiological functions, including endothelial cell function. This study explores the distinct roles of two essential autophagy regulators, autophagy-related gene 5 (ATG5) and ATG7, in modulating TGF-{beta} signalling and endothelial function. To this end, ATG5 and ATG7 were selectively silenced in endothelial cells, with knockdown efficiency confirmed via RT-qPCR and Western blot analyses. Gene expression profiling revealed differential regulation of TGF-{beta} signalling components following ATG5 or ATG7 silencing. Functional assays demonstrated that ATG5 knockdown enhanced endothelial cell proliferation and migration, whereas ATG7 silencing produced distinct, less pronounced effects. Furthermore, downstream effectors of the TGF-{beta} pathway exhibited gene-specific modulation, underscoring divergent roles of ATG5 and ATG7 in this signalling cascade. Collectively, these findings highlight the non-redundant functions of ATG5 and ATG7 in coordinating TGF-{beta} signalling pathways, offering new insights into their contribution to endothelial physiology and potential as therapeutic targets in vascular pathologies.

cell biology↗

BRCA2-deficiency Causes Global Transcriptomic Alterations in Endothelial Cells

This study aims to explore the alterations in gene expression following BRCA2 loss in endothelial cells and to investigate the potential contribution of BRCA2 in regulating the endothelial protein-coding transcriptome. Cultured human umbilical vein endothelial cells (HUVECs) were transfected with siBRCA2 or non-targeting scrambled RNA as control. Total RNA was extracted and used for RNA sequencing following purification and quality check. Transcription profiles and differentially expressed genes (DEGs) were identified. Enrichment of functions and signaling pathways analysis were performed based on Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. In total, 3196 DEGs with (cut-off 1.5-fold) were identified, of which 1852 genes were upregulated, and 1344 genes were downregulated in BRCA2-deficient HUVECs. Cell cycle exit and neuronal differentiation 1 (CEND1) and mannose receptor C-type 1 (MRC1) were the most significantly up- and downregulated genes, respectively, in BRCA2-deficient endothelial cells. The expression of top DEGs were further validated by RT-qPCR in HUVECs and in human coronary artery endothelial cells. The GO and KEGG analysis suggest that cell adhesion molecules and cytokine-cytokine receptor interaction may play an important role in BRCA2-deficient endothelial cells. Overall, these findings contribute to a greater understanding of the mechanisms involved in BRCA2-mediated endothelial function.

genomics↗

The effect of fatty acid binding protein-3 exposure on endothelial transcriptomics

BackgroundFatty acid binding protein-3 (FABP3) is released in circulation following myocardial infarction and an increased level of circulatory FABP3 is also reported in peripheral artery disease patients exposing endothelial cells to higher level of FABP3. Recently, loss of endothelial FABP3 was shown to protect endothelial cells against inflammation-induced endothelial dysfunction, however, the effect of FABP3-exposure on endothelial cells is unknown. Accordingly, to study the effect of FABP3-exposure on endothelial cells we performed transcriptomic profiling following recombinant human FABP3 (rhFABP3) treatment to endothelial cells. MethodCultured human endothelial cells were treated with either vehicle or rhFABP3 (50 ng/ml, 6h) and then RNA-sequencing was performed. Gene expression analysis followed by gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were performed to identify differentially expressed genes and affected cellular functions and pathways. ResultsDifferential gene expression analysis revealed kinesin family member 26b (KIF26B) as the most upregulated and survival of motor neuron 2 (SMN2) as the most downregulated genes in rhFABP3-treated in comparison to vehicle-treated endothelial cells. Most of the differentially expressed genes were associated with endothelial cell motility, immune response, and angiogenesis. GO and KEGG analyses indicated that rhFABP3 exposure impacts several crucial pathways, prominently "Regulation of leukocyte mediated cytotoxicity" and "Natural killer cell mediated cytotoxicity", suggesting its involvement in endothelial cell physiology and response mechanisms to cardiovascular stress. ConclusionThis is the first study evaluating rhFABP3-induced transcriptomics in human endothelial cells. Our data reveal novel genes and pathways affected by FABP3-exposure to endothelial cells. Further research is necessary to validate these findings and fully understand FABP3s role in endothelial biology and in cardiovascular diseases like myocardial infarction and peripheral artery disease.

biochemistry↗

Screening of Natural Plant Extracts for Antimicrobial Activity Against Streptobacillus moniliformis

The rise of antimicrobial resistance has driven the search for alternative antibacterial agents, including plant-based compounds. This study evaluates the antimicrobial potential of selected herbal extracts against Streptobacillus moniliformis using the agar well diffusion method. The tested extracts included Tulsi leaves (Ocimum sanctum), Neem leaves (Azadirachta indica), Bael leaves (Aegle marmelos), Adarak peels (Zingiber officinalef), Moringa seeds and leaves (Moringa oleifera), Dalchini (Cinnamomum verum), Lemon/Orange peels (Citrus limon and Citrus sinensis), and Ginger peels (Zingiber officinale). Among these, Bael leaves and Lemon peels demonstrated significant antibacterial activity, forming distinct zones of inhibition. In contrast, Neem and Moringa extracts did not inhibit bacterial growth. The observed antimicrobial activity is likely due to the presence of bioactive compounds such as flavonoids, tannins, and essential oils, which may disrupt bacterial cell walls and metabolic processes. Notably, S. moniliformis exhibited limited survival in culture, while other bacterial strains showed minimal resistance. These findings suggest that certain herbal extracts, particularly Bael leaves and Lemon peels, may serve as natural antimicrobial agents against S. moniliformis. Further studies are required to isolate and characterize the active compounds responsible for this antibacterial activity to explore their potential in developing alternative antimicrobial therapies.

microbiology↗

THE CORRELATION BETWEEN REDOX ACTIVITY AND ANTIMICROBIAL PROPERTIES OF PYOCYANIN FROM Pseudomonas aeruginosa

Pseudomonas aeruginosa, a metabolically versatile gram-negative bacterium, has garnered attention for its ability to thrive in diverse environments and its capacity to produce pyocyanin, a secondary metabolite with multifunctional properties. Pyocyanin, a redox-active phenazine compound, plays a critical role in mediating the ecological competitiveness of P. aeruginosa through its antimicrobial, biofilm-modulating, and reactive oxygen species (ROS)-generating activities. Beyond its contributions to bacterial virulence, pyocyanin demonstrates significant potential in various industrial and biomedical applications due to its redox properties and ability to function under diverse environmental conditions. This study investigates the electrochemical behaviour and pH-dependent antimicrobial activity of pyocyanin to evaluate its applicability in environmental and medical fields. Soil-derived P. aeruginosa isolates were cultured for pyocyanin (PYO) production, and the pigment was characterized using UV-visible spectroscopy to get its spectral integrity. Electrochemical analysis through cyclic voltammetry revealed enhanced redox activity in acidic environments and stable functionality in alkaline conditions. Antimicrobial assays demonstrated that pyocyanin exhibited optimal activity at neutral to slightly alkaline pH, effectively inhibiting bacterial and fungal growth, while extreme acidic conditions reduced its efficacy. The findings highlight pyocyanins versatility as both a redox mediator and an antimicrobial agent. In medical contexts, its pH-sensitive activity aligns well with physiological conditions, offering promise for combating multidrug-resistant pathogens. Future optimization of pyocyanin biosynthesis, particularly through cost-effective and scalable methods, could unlock its full potential in biotechnological and therapeutic innovations.

microbiology↗

The sodium channel SCN2A regulates cortical excitatory and inhibitory neurogenesis

Voltage-gated sodium channels regulate neuronal excitability and synaptic transmission in the postnatal and adult brain. The gene SCN2A, encoding the sodium channel Nav1.2, regulates synaptic development and variants in SCN2A are associated with autism spectrum disorders (ASD) and a broad spectrum of epilepsy phenotypes, including early-onset developmental and epileptic encephalopathies. The expression pattern of SCN2A begins during prenatal cortical development, prior to the onset of synaptic transmission, but it is unknown whether SCN2A regulates early cortical development through mechanisms independent of synaptic transmission. Here we reveal that isogenic and ASD patient-derived human forebrain organoids modelling a loss of SCN2A function display impaired excitatory and inhibitory neurogenesis, leading to a developmental imbalance. Unexpectedly, we find precocious generation of cortical inhibitory neurons is driven by elevated Sonic hedgehog signaling and is reversible through pharmacological inhibition. Functionally, these developmental phenotypes arise due to Nav1.2-dependent sodium channel dysfunction and reduced action potential generation, leading to abnormal neuronal network activity. Our results identify a mechanism for cortical excitatory and inhibitory neurogenesis involving SCN2A, and reveal that early neurogenesis deficits precede postnatal neural circuit dysfunction in SCN2A-associated disorders.

neuroscience↗

Chloroquine Up-regulates Expression of SARS-CoV-2 receptor Angiotensin Converting Enzyme-2 in Endothelial Cells

Background and PurposeThe novel severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) posed a serious threat to global public health. Hydroxychloroquine (HCQ), which is a derivative of Chloroquine (CQ), was a WHO-recommended drug to treat COVID-19 with mixed effects. The purpose of the present study is to evaluate the plausible mechanisms of HCQ actions behind its observed mixed effect. Key ResultsWe demonstrate that CQ-treatment significantly up-regulates mesenchymal markers and SARS-CoV-2 receptor ACE2 in cultured endothelial cells. Conclusions & ImplicationsThe detrimental effect of HCQ in seriously ill COVID-19 patients might be due to CQ-induced increased expression of endothelial ACE2 exacerbating the severity of SARS-CoV-2 infection. Our study warrants further investigation in animal models and humans and caution while prescribing HCQ to patients with an impaired renin-angiotensin-aldosterone system, such as in hypertension, cardiovascular diseases, or chronic kidney disease; particularly with ACE-inhibitors or statin therapy.

cell biology↗

Natural Product Screen Identifies Asiatic Acid Targeting Mutant p53

Mutations in p53 are common in different cancer types and are reported to protect cancer via different mechanisms. R175H, R248Q, and R273H are the hotspot mutations of p53 and are suggested to increase aerobic glycolysis in cancer cells. Cancer cells rely mostly upon aerobic glycolysis, so it will be interesting to target these three p53 mutants for designing alternative cancer therapy. The class of compounds studied for their potential to target energy metabolism of cancer cells is called mitocans. The current study is an approach to explore if these selected 3 mutants of p53 may act as suitable target(s) for natural mitocans. Hereby, we selected 60 phytocompounds altogether from Andrographis paniculata and Centella asiatica and docked against all three p53 mutants, using Autodock vina. 11 compounds were sorted based on their binding energies and drug-like properties, and toxicity levels prediction showed asiatic acid to be the most significant. As asiatic acid was observed to significantly bind with R248Q only, R248Q-Asiatic acid was identified for molecular dynamics simulation using GROMACS which showed significant interactions. In conclusion, mutant R248Q was observed to be the best target for asiatic acid, though additional in-vitro experiments are important to validate the findings of this study.

cancer biology↗

Ischemia/Reperfusion Injury and Oxidative Stress Impair Cardiac Desmin Proteostasis

BackgroundThough acute mortality by myocardial infarction (MI) has declined in past decades, MI still represents one of the leading causes of heart failure (HF) development. We recently demonstrated the accumulation of toxic desmin aggregates in patients with HF of ischemic origin. Since desmin aggregates are toxic for the heart we aimed to test whether their formation can be induced by oxidative stress as a proxy for reperfusion injury, as well as addressing the effects of therapeutic strategies aimed at reducing desmin aggregation with cardiac oxidative stress. Methods and ResultsWe demonstrate here that oxidative stress is able to induce desmin aggregation, acutely, in a cell-specific and dose-dependent fashion. We also show that elevation of O-linked {beta}-N-acetylglucosamine (O-GlcNAc) prior to or after oxidative stress reduces the formation of toxic desmin aggregates and its pro-aggregating desmin post-translational modifications (PTM). In addition, we show for the first time a role for the transmembrane protease serine 13 (TMPRSS13) with desmin cleavage in response to oxidative stress while desmins single cysteine plays a protective role from I/R injury, which is independent of gain or loss of desmin function. ConclusionsThe proliferation of desmin PTM-forms (i.e., proteoforms) and its aggregation hallmark acute and chronic cardiac stress and result in both loss of and gain of desmin function. We report here two novel mechanisms that could be targeted for therapy to preserve desmin homeostasis and cardiac function in the acute settings of oxidative stress and reperfusion injury.

biochemistry↗

TAOK2β represses translation via phosphorylation of eEF2 and ameliorates exaggerated protein synthesis in a mouse model of 16p11.2 microdeletion-driven autism

Microdeletions in the 16p11.2 region of the human genome are frequently associated with autism spectrum disorders (ASDs), but how these genomic rearrangements cause ASD remains unclear. Here, we reveal that TAOK2{beta}, a protein isoform encoded by the human TAOK2 gene located in the 16p11.2 locus, regulates mRNA translation. To identify key functional interaction partners of TAOK2{beta}, we performed proteomic screening from Neuro-2a (N2a) cells, mouse cortices, and cultured neurons. This revealed translation factors as a major class of enriched interacting proteins. Consistently, TAOK2{beta} is present in mouse cortical polyribosomes and cortices from Taok2 knockout mice show increased ribosome density on mRNAs and enhanced protein synthesis. Several lines of evidence support an effect of TAOK2{beta} on translation elongation via phosphorylation of eukaryotic elongation factor (eEF2). TAOK2 can directly phosphorylate eEF2 on Threonine 56 and this phosphorylation is reduced in cortices from Taok2 knockout mice. TAOK2{beta} WT overexpression increased eEF2 phosphorylation levels and reduced protein synthesis, whereas a kinase-dead allele of TAOK2{beta} showed opposite effects. Finally, we show that cortices from the mouse model of the human 16p11.2 microdeletion have increased polysome/monosome (P/M) ratios and protein synthesis, phenocopying Taok2 loss of function. Importantly, defective translation phenotypes observed in the mouse 16p11.2 microdeletion model of ASD could be normalized either by reintroducing Taok2 in vivo or by delivering TAOK2{beta} to cortical neurons derived from 16p11.2 microdeletion mice. Our results uncover a critical role of TAOK2{beta} as a regulator of protein synthesis and support the idea that translational control is a common endpoint of ASD-associated signaling pathways.

neuroscience↗

Neuron-specific protein network mapping of autism risk genes identifies shared biological mechanisms and disease relevant pathologies

Manuscript summaryThere are hundreds of risk genes associated with autism spectrum disorder (ASD), but signaling networks at the protein level remain unexplored. We use neuron-specific proximity-labeling proteomics (BioID) to identify protein-protein interaction (PPI) networks for 41 ASD-risk genes. Neuron-specific PPI networks, including synaptic transmission proteins, are disrupted by de novo missense variants. The PPI network map reveals convergent pathways, including mitochondrial/metabolic processes, Wnt signaling, and MAPK signaling. CRISPR knockout reveal an association between mitochondrial activity and ASD-risk genes. The PPI network shows an enrichment of 112 additional ASD-risk genes and differentially expressed genes from post-mortem ASD patients. Clustering of risk genes based on PPI networks identifies gene groups corresponding to clinical behavior score severity. Our data reveal that cell type-specific PPI networks can identify individual and convergent ASD signaling networks, provide a method to assess patient variants, and reveal biological insight into disease mechanisms and sub-cohorts of ASD.

neuroscience↗

Impaired OTUD7A-dependent Ankyrin regulation mediates neuronal dysfunction in mouse and human models of the 15q13.3 microdeletion syndrome

Copy number variations (CNV) are associated with psychiatric and neurodevelopmental disorders (NDDs), and most, including the recurrent 15q13.3 microdeletion disorder, have unknown disease mechanisms. We used a heterozygous 15q13.3 microdeletion mouse model and patient iPSC-derived neurons to reveal developmental defects in neuronal maturation and network activity. To identify the underlying molecular dysfunction, we developed a neuron-specific proximity-labeling proteomics (BioID2) pipeline, combined with patient mutations, to target the 15q13.3 CNV genetic driver OTUD7A. OTUD7A is an emerging independent NDD risk gene with no known function in the brain, but has putative deubiquitinase (DUB) function. The OTUD7A protein-protein interaction (PPI) network revealed interactions with synaptic, axonal, and cytoskeletal proteins and was enriched for known ASD and epilepsy risk genes. The interactions between OTUD7A and the NDD risk genes Ankyrin-G (Ank3) and Ankyrin-B (Ank2) were disrupted by an epilepsy-associated OTUD7A L233F variant. Further investigation of Ankyrin-G in mouse and human 15q13.3 microdeletion and OTUD7AL233F/L233F models revealed protein instability, increased polyubiquitination, and decreased levels in the axon initial segment (AIS), while structured illumination microscopy identified reduced Ankyrin-G nanodomains in dendritic spines. Functional analysis of human 15q13.3 microdeletion and OTUD7AL233F/L233F models revealed shared and distinct impairments to axonal growth and intrinsic excitability. Importantly, restoring OTUD7A or Ankyrin-G expression in 15q13.3 microdeletion neurons led to a reversal of abnormalities. These data reveal a critical OTUD7A-Ankyrin pathway in neuronal development, which is impaired in the 15q13.3 microdeletion syndrome, leading to neuronal dysfunction. Further, our study highlights the utility of targeting CNV genes using cell-type specific proteomics to identify shared and unexplored disease mechanisms across NDDs.

neuroscience↗

Disruption of the autism-associated gene SCN2A alters synaptic development and neuronal signaling in patient iPSC-glutamatergic neurons

SCN2A is an autism spectrum disorder (ASD) risk gene and encodes a voltage-gated sodium channel. However, the impact of autism-associated SCN2A de novo variants on human neuron development is unknown. We studied SCN2A using isogenic SCN2A-/- induced pluripotent stem cells (iPSCs), and patient-derived iPSCs harboring a p.R607* or a C-terminal p.G1744* de novo truncating variant. We used Neurogenin2 to generate excitatory glutamatergic neurons and found that SCN2A+/p.R607* and SCN2A-/- neurons displayed a reduction in synapse formation and excitatory synaptic activity using multielectrode arrays and electrophysiology. However, the p.G1744* variant, which leads to early-onset seizures in addition to ASD, altered action-potential dynamics but not synaptic activity. Proteomic and functional analysis of SCN2A+/p.R607* neurons revealed defects in neuronal morphology and bioenergetic pathways, which were not present in SCN2A+/p.G1744* neurons. Our study reveals that SCN2A de novo variants can have differential impact on human neuron function and signaling. HIGHTLIGHTS- Isogenic SCN2A-/- neurons display intrinsic hyperexcitability and impaired excitatory synapse function - SCN2A+/p.R607* variant reduces excitatory synapse function in patient neurons - C-terminal SCN2A+/p.G1744* variant enhances action potential properties but not synaptic transmission in patient neurons - SCN2A+/p.R607* variant display impacts on morphological and bioenergetic signaling networks through proteomic and functional analysis eTOC- Brown et al. examined Autism-associated SCN2A variants using patient-derived iPSC NGN2-neurons. They discover that genetic variants differentially impact neuronal development and synaptic function, and highlight neuronal and bioenergetic signaling networks underlying SCN2A loss-of-function.

neuroscience↗

Deep learning models for identification of splice junctions across species

Deep learning models like convolutional neural networks (CNN) and recurrent neural networks (RNN) have been frequently used to identify splice sites from genome sequences. Most of the deep learning applications identify splice sites from a single species. Furthermore, the models generally identify and interpret only the canonical splice sites. However, a model capable of identifying both canonical and non-canonical splice sites from multiple species with comparable accuracy is more generalizable and robust. We choose some state-of-the-art CNN and RNN models and compare their performances in identifying novel canonical and non-canonical splice sites in homo sapiens, mus musculus, and drosophila melanogaster. The RNN-based model named SpliceViNCI outperforms its counterparts in identifying splice sites from multiple species as well as on unseen species. SpliceViNCI maintains its performance when trained with imbalanced data making it more robust. We observe that all the models perform better when trained with more than one species. SpliceViNCI outperforms the counterparts when trained with such an augmented dataset. We further extract and compare the features learned by SpliceViNCI when trained with single and multiple species. We validate the extracted features with knowledge from the literature.

bioinformatics↗

The MAGOH paralogs - MAGOH, MAGOHB and their multiple isoforms

A central processing event in eukaryotic gene expression is splicing. Concurrent with splicing, the core-EJC proteins, eIF4A3 and RBM8A-MAGOH heterodimer are deposited 24 bases upstream of newly formed exon-exon junctions. One of the core-EJC proteins, MAGOH contains a paralog MAGOHB, and this paralog pair is conserved across vertebrates. Upon analysis of the splice variants of MAGOH-paralogs, we have found the presence of alternate protein isoforms which are also evolutionarily conserved. Further, comparison of the amino acid sequence of the principal and alternate protein isoforms has revealed absence of key amino acid residues in the alternate isoforms. The conservation of principal and alternate isoforms correlates to the importance of MAGOH and MAGOHB across vertebrates.

molecular biology↗

SpliceViNCI: Visualizing the splicing of non-canonical introns through recurrent neural networks

Most of the current computational models for splice junction prediction are based on the identification of canonical splice junctions. However, it is observed that the junctions lacking the consensus dimers GT and AG also undergo splicing. Identification of such splice junctions, called the non-canonical splice junctions, is also essentially important for a comprehensive understanding of the splicing phenomenon. This work focuses on the identification of non-canonical splice junctions through the application of a bidirectional long short-term memory (BLSTM) network. Furthermore, we apply a back-propagation based (integrated gradient) and a perturbation based (occlusion) visualization techniques to extract the non-canonical splicing features learned by the model. The features obtained are validated with the existing knowledge from the literature. Integrated gradient extracts features that comprise contiguous nucleotides, whereas occlusion extracts features that are individual nucleotides distributed across the sequence.

bioinformatics↗