Biology subjects

Basnet, S.

Publications and source records attributed to Basnet, S..

9 recordsLinked to original sources

bioRxiv2026

TransBind2: Improving Transcription Factor-DNA Binding Prediction with Multimodal Data and Bidirectional Cross Attention

Accurate genome-wide prediction of transcription factor (TF)-DNA binding remains challenging because many models focus mainly on DNA sequence and overlook chromatin context and TF structure. We previously developed TransBind, a protein-aware model that combines TF and DNA representations through cross-attention. Here, we introduce TransBind2, which improves on TransBind in several ways. It incorporates DNase-seq accessibility and genome mappability tracks as additional input, uses a biomodal protein language model (ProstT5) to capture both TF sequence and structure, and applies bidirectional cross-attention so DNA and protein features can refine each other. We also frame prediction as binary classification of individual triplets, allowing the model to generalize to new TFs and cell types. Across 690 human ChIP-seq experiments covering 161 TFs and 91 cell types, TransBind2 achieves a macro AUROC of 0.9648 and AUPR of 0.4215, outperforming TransBind and other baselines, with a [≥]12.67% relative AUPR gain. The model trained on human data also performs well in cross-species zero-shot prediction on mouse data. Saliency analysis shows that it can identify TF-binding peaks with a median error of 12-38 base pairs (bps) despite being trained on window-level labels. Ablation studies further show that TF structure, chromatin accessibility, and bidirectional attention each improve performance. Overall, these results show that combining TF structure with chromatin context leads to more accurate and generalizable TF-DNA binding predictions.

bioinformatics↗

bioRxiv2026

Less is more? Faster growth is associated with lower ectomycorrhizal diversity in mature Picea glauca at the Alaskan treelines

Root-associated fungal (RAF) communities can influence tree nutrition and performance, yet their environmental drivers and relationship with tree growth remain poorly understood, particularly near treelines. We characterized RAF communities, with a focus on ectomycorrhizal (ECM) fungi, associated with fine roots of white spruce (Picea glauca [Moench] Voss) across paired forest and treeline plots representing two elevational and one drought-limited treeline ecotones in Alaska. Using ITS2 DNA metabarcoding, we assessed fungal community composition and alpha diversity and related these metrics to basal area increment (BAI) over 5-30 years for individual trees. Sampling site was the strongest predictor of RAF community composition, explaining 19.6% of variation, followed by soil pH (11.7%), whereas habitat explained only 2.1%. Treeline effects were context-dependent: RAF composition differed between forest and treeline only in the Alaska Range, alpha diversity was lower at the treeline in Interior Alaska, and ECM relative abundance was higher at the Brooks Range treeline. In contrast, RAF community composition did not differ between fast- and slow-growing trees within sites. Interestingly, alpha diversity indices were significantly associated with BAI over the previous 5, 10 and 15 years, with lower diversity associated with faster tree growth. These relationships weakened with increasing BAI averaging period. Our results indicate that, in mature P. glauca, tree growth is associated with lower RAF and ECM diversity but not with distinct fungal community composition, suggesting that growth may be linked to greater dominance by a subset of fungal partners rather than higher fungal diversity.

ecology↗

bioRxiv2026

TFBindFormer:A Cross-Attention Transformer for Transcription Factor--DNA Binding Prediction

Transcription factors (TFs) are central regulators of gene expression, and their selective recognition of genomic DNA underlies various biological processes. Experimental profiling of TF-DNA interactions using chromatin immunoprecipitation followed by sequencing (ChIP-seq) provides high-resolution maps of in vivo TF-DNA binding but remains costly, labor-intensive, and inherently low-throughput, limiting their scalability across different transcription factors, cell types, and regulatory conditions. Computational modeling therefore plays an essential role in inferring TF-DNA interactions at genome scale. However, most existing computational models rely solely on DNA sequence and chromatin features to predict TF-DNA binding, neglecting TF-specific protein information. This omission limits their ability to capture protein-dependent binding specificity. Here, we present TFBindFormer, a hybrid cross-attention transformer that explicitly integrates genomic DNA features with TF-specific representations derived from protein sequences and structures. By modeling protein-conditioned, position-specific TF-DNA interactions, TFBindFormer enables direct learning of molecular determinants underlying DNA recognition. Evaluated across hundreds of cell-type-specific TFs and hundreds of millions of genome-wide DNA bins, TFBindFormer consistently outperforms DNA-only baselines, achieving substantial gains in both area under precision-recall curve (AUPRC) and area under receiver operating characteristic curve (AUROC). Together, these results demonstrate that integrating TF and DNA features via cross-attention enables TFBindFormer to serve as an effective and scalable framework for large-scale TF-DNA binding prediction.

bioinformatics↗

bioRxiv2025

Integrating Protein and DNA Embeddings for Improving Genome-Wide Transcription Factor Binding Site Prediction

Transcription factors (TFs) regulate gene expression by binding to specific DNA sites on genome, making accurate TF binding site prediction critical for understanding gene regulation and downstream phenotypes. Current deep learning methods use only DNA-related information to predict TF binding sites, ignoring the fact that different TF protein sequences and structures recognize distinct DNA patterns. Not leveraging TF information not only limits prediction accuracy but also makes the methods not generalizable to predicting binding sites of new TFs that do not exist in the traning data. Here, we present TransBind, a protein-aware deep learning architecture that integrates DNA sequence information with protein embeddings containing both sequence and structural information derived from a protein language model pretrained on DNA-binding proteins, to improve TF binding site prediction. Through the cross-attention, a TF embedding selectively attends to genomic regions according to its unique binding properties. Evaluated on the data of 690 ChIP-seq experiments spanning 161 TFs across 91 human cell types, TransBind achieves an AUROC of 0.950 and AUPR of 0.371--representing a [≥]11.3% relative AUPR improvement over state-of-the-art methods including TBiNet, DanQ, and DeepSEA. The model outperformed existing methods in [≥]97.1% of TF-cell type combinations. It also recovered 160 known TF binding motifs in the JASPAR database, providing the biological interpretability of the model. Moreover, the approach enables zero-shot prediction for unseen TFs, demonstrating its potential of generalizing to new, poorly characterized TFs. The source code of TransBind is available at https://github.com/jianlin-cheng/TransBind.

bioinformatics↗

bioRxiv2024

Developmental and physiological impacts of pathogenic human huntingtin protein in the nervous system

Huntingtons Disease (HD) is a neurodegenerative disorder, part of the nine identified inherited polyglutamine (polyQ) diseases. Most commonly, HD pathophysiology manifests in middle-aged adults with symptoms including progressive loss of motor control, cognitive decline, and psychiatric disturbances. Associated with the pathophysiology of HD is the formation of insoluble fragments of the huntingtin protein (htt) that tend to aggregate in the nucleus and cytoplasm of neurons. To track both the intracellular progression of the aggregation phenotype as well as the physiological deficits associated with mutant htt, two constructs of human HTT were expressed with varying polyQ lengths, non-pathogenic-htt (Q15, NP-htt) and pathogenic-htt (Q138, P-htt), with an N-terminal RFP tag for in vivo visualization. P-htt aggregates accumulate in the ventral nerve cord cell bodies as early as 24 hours post hatching and significant aggregates form in the segmental nerve branches at 48 hours post hatching. Organelle trafficking up-and downstream of aggregates formed in motor neurons showed severe deficits in trafficking dynamics. To explore putative downstream deficits of htt aggregation, ultrastructural changes of presynaptic motor neurons and muscles were assessed, but no significant effects were observed. However, the force and kinetics of muscle contractions were severely affected in P-htt animals, reminiscent of human chorea. Reduced muscle force production translated to altered locomotory behavior. A novel HD aggregation model was established to track htt aggregation throughout adulthood in the wing, showing similar aggregation patterns with larvae. Expressing P-htt in the adult nervous system resulted in significantly reduced lifespan, which could be partially rescued by feeding flies the mTOR inhibitor rapamycin. These findings advance our understanding of htt aggregate progression as well the downstream physiological impacts on the nervous system and peripheral tissues.

neuroscience↗

bioRxiv2024

Proteasome inhibition induces DNA methylation alteration by attenuating the synthesis of DNA methyltransferase 1 and 3B in colorectal cancer

Proteasome is an essential organelle in guarding cellular protein homeostasis. Here, we report that inhibition of proteasome leads to alterations in DNA methylation patterns in colorectal cancer (CRC) by surpressing the synthesis of DNA methyltransferases (DNMTs). We found that treating CRC cells with proteasome inhibitors results in attenuated translation of DNMT1 and DNMT3B, mediated by the inactivation of AKT and mammalian target of rapamycin (mTOR), which is dependent on the accumulation of p300, an acetyltransferase that inhibits AKT through acetylation modification. Furthermore, we demonstrated that downregulation of DNMT1 and DNMT3B confers protection against proteasome inhibitor treatment, potentially through reprogramming the transcriptome of CRC cells, highlighting the significant role of DNMTs in response to disruptions in protein homeostasis. Interestingly, the proteasome inhibitor-induced downregulation of DNMT1 and DNMT3B appears to be CRC specific, notwithstanding the underlying mechanism remains unclear. Altogether, our findings reveal an epigenetic effect of proteasome on DNA methylation in CRC through its regulation of DNA methyltransferase synthesis.

cancer biology↗

bioRxiv2022

Identification of potential human pancreatic α-amylase inhibitors from natural products by molecular docking, MM/GBSA calculations, MD simulations, and ADMET analysis

Human pancreatic -amylase (HPA), which works as a catalyst for carbohydrate hydrolysis, is one of the viable targets to control type 2 diabetes. The inhibition of -amylase lowers blood glucose levels and helps to alleviate hyperglycemia complications. Herein, we systematically screened the potential HPA inhibitors from an in-house library of natural products by molecular modeling. The modeling encompasses molecular docking, MM/GBSA binding energy calculations, MD simulations, and ADMET analysis. This research identified newboulaside B, newboulaside A, quercetin-3-O-{beta}-glucoside, and sasastilboside A as the top four potential HPA inhibitors from the library of natural products, whose Glide docking scores and MM/GBSA binding energies range from -9.191 to -11.366 kcal/mol and -19.38 to -50.29 kcal/mol, respectively. Based on the simulation, among them, newboulaside B was found as the best HPA inhibitor. Throughout the simulation, with the deviation of 3[A] (acarbose = 3[A]), it interacted with ASP356, ASP300, ASP197, THR163, ARG161, ASP147, ALA106, and GLN63 via hydrogen bonding. Additionally, the comprehensive ADMET analysis revealed that it has good pharmacokinetic properties having not acutely toxic, moderately bioavailable, and non-inhibitor nature toward cytochrome P450. All the results suggest that newboulaside B might be a promising candidate for drug discovery against type 2 diabetes. Graphical Abstract O_FIG_DISPLAY_L [Figure 1] M_FIG_DISPLAY C_FIG_DISPLAY

bioinformatics↗

bioRxiv2020

In Silico Analysis of Secondary Metabolites that Modulates Enzymes of Cholesterol Target

Hypercholesterolemia has posed a serious threat of heart diseases and stroke worldwide. Xanthine oxidase (XO), the rate-limiting enzyme in uric acid biosynthesis, is regarded as the root of reactive oxygen species (ROS) that generates atherosclerosis and cholesterol crystals. {beta}-Hydroxy {beta}-methylglutaryl-coenzyme A reductase (HMGR) is a rate-limiting enzyme in cholesterol biosynthesis. Although some commercially available enzyme inhibiting drugs have effectively reduced the cholesterol level, most of them have failed to meet the requirements of being apt drug candidates. Here, we have carried out an in-silico analysis of secondary metabolites that have already shown good inhibitory activity against XO and HMGR. Out of 118 secondary metabolites reviewed, sixteen molecules inhibiting XO and HMGR were taken based on IC50 values reported in vitro assays. Further, receptor-based virtual screening was carried out against secondary metabolites using GOLD Protein-Ligand Docking Software, combined with subsequent post-docking, to study the binding affinities of ligands to the enzymes. In-Silico ADMET analysis was carried out to study their pharmacokinetic properties, followed by toxicity prediction through ProTox-II. The molecular docking of amentoflavone (1) (GOLD score 70.54), and ganomycin I (9) (GOLD score 59.61) evinced that the drug has effectively bind at the competitive site of XO and HMGR, respectively. Besides, 6-paradol (3) and selgin (4) could be potential drug candidates to inhibit XO. Likewise, n-octadecanyl-O--D-glucopyranosyl(6[->]1")-O--D-glucopyranoside (10) could be potential drug candidates to maintain serum cholesterol. In-silico ADMET analysis showed that the sixteen metabolites were optimal within the categorical range in comparison to commercially available XO and HMGR inhibitors, respectively. Toxicity analysis through Protox-II revealed that 6-gingerol (2), ganoleucoin K (11), and ganoleucoin Z (12) are toxic for human use. This computational analysis supports earlier experimental evidence towards the inhibition of XO and HMGR by natural products. Further study is necessary to explore the clinical efficacy of these secondary molecules, which might be alternatives for the treatment of hypercholesterolemia. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=163 HEIGHT=200 SRC="FIGDIR/small/219998v2_ufig1.gif" ALT="Figure 1"> View larger version (59K): org.highwire.dtl.DTLVardef@48d66borg.highwire.dtl.DTLVardef@1f51741org.highwire.dtl.DTLVardef@1c9d87dorg.highwire.dtl.DTLVardef@a41e70_HPS_FORMAT_FIGEXP M_FIG C_FIG

biochemistry↗

bioRxiv2018

Pre-existing Technological Core and Roots for the CRISPR Breakthrough

This paper applies objective methods to explore the technological origins of the widely acclaimed CRISPR breakthrough in the technological domain of genome engineering. Previously developed patent search techniques are first used to recover a set of patents that well-represent the genome editing domain before CRISPR. Main paths are then determined from the citation network associated with this patent set allowing identification of the three major knowledge trajectories. The most significant of these trajectories for CRISPR involves the core of genome editing with less significant trajectories involving cloning and endonuclease specific developments. The major patents on the core trajectory are consistent with qualitative expert knowledge of the topical area. A second set of patents that we call the CRISPR roots are obtained by finding the patents directly cited by the recent CRISPR patents along with patents cited by that set of patents. We find that the CRISPR roots contain 8 key patents from the genome engineering main path associated with restriction endonucleases and the expected strong connection of CRISPR to prior genome editing technology such as Zn finger nucleases. Nonetheless, analysis of the full CRISPR roots shows that a very wide array of technological knowledge beyond genome engineering has contributed to achieving the CRISPR breakthrough. Such breadth in origins is not surprising since \"spillover\" is generally perceived as important and previous qualitative studies of CRISPR have shown not only technological breadth in origins but scientific breadth as well. In addition, we find that the estimated rate of functional performance improvement of the CRISPR roots set is about 9% per year compared to the genome engineering set ({small tilde}4 % per year). These estimates indicate below average rates of improvement and may indicate that CRISPR (and perhaps yet undiscovered) genome engineering developments could evolve in effectiveness over an upcoming long rather than short time period.

bioengineering↗