Biology subjects

Raghava, G. P. S.

Publications and source records attributed to Raghava, G. P. S..

At least 37 records · Page 2Linked to original sources

bioRxiv2025

AntiCP3: Prediction of Anticancer Proteins Using Evolutionary Information from Protein Language Models

A number of computational methods have been developed in the past for predicting anticancer peptides, including AntiCP and AntiCP2 from our group. While these tools have been widely used by the scientific community, they are not suitable for predicting anticancer proteins. In this study, we present AntiCP3, the first dedicated method for the prediction of anticancer proteins. All models were trained using five-fold cross-validation and evaluated on an independent dataset not used during training. Our initial analysis revealed distinct compositional differences between anticancer peptides and proteins, justifying the need for a separate prediction framework. We first implemented similarity-based approaches, which yielded moderate performance. Subsequently, we developed machine learning and deep learning models using conventional protein features, achieving a maximum AUC of 0.72. The performance improved to an AUC of 0.79 with the incorporation of evolutionary information through PSSM profiles. Further enhancement was observed when embeddings from a fine-tuned protein language model ESM-t33 were used, leading to a best AUC of 0.90. Finally, a hybrid approach combining BLAST with our machine learning model achieved an AUC of 0.91. To facilitate the scientific community, we have implemented AntiCP3 as both a web server and standalone software for the prediction of anticancer proteins (https://webs.iiitd.edu.in/raghava/anticp3/). We have also deployed our model at hugging face https://huggingface.co/raghavagps-group/anticp3. Highlights Existing methods have been developed for predicting anticancer peptides. AntiCP3 is specifically optimised for predicting anticancer proteins. PSI-BLAST used to obtain evolutionary information in form of PSSM profile. Hybrid methods developed using alignment free and alignment based approach. A web server and standalone tool have been developed to assist the community.

bioinformatics↗

bioRxiv2025

IL4Pred2: Prediction of Interleukin-4 Inducing Peptides in Human and Mouse

In 2013, our group developed IL4pred, a host-independent method for predicting interleukin-4 (IL-4) inducing peptides, which has been widely used by the scientific community. In this study, we present a second-generation method, IL4Pred2, which is a host-specific approach designed to predict IL-4 inducing peptides separately for human and mouse hosts. All models were trained, tested, and benchmarked on experimentally validated data obtained from the IEDB. We employed a wide range of state-of-the-art techniques for prediction, including similarity-based approaches, machine learning, deep learning methods, and large language models. Our best model achieved highest AUC 0.80 with MCC 0.45 for human and AUC 0.82 with MCC 0.50 for mouse on independent set of main datasets. All models were trained, test and optimized on training dataset. We validate our final model on an independent dataset which is not used in training or hyperparameter optimization of models. In this study, we developed models on three types of datasets called Main, Alternate1 and Alternate2 for predicting IL-4 inducing peptides. This abstract show performance of our models on Main dataset, performance of models on other datasets have been discussed in manuscript. One of the major objectives of this study is to facilitate research community in the area of immunotherapy and vaccine development. Thus, we developed, a web server and standalone software IL4pred2 for predicting, designing and scanning IL-4 inducing peptides in proteins (https://webs.iiitd.edu.in/raghava/il4pred2/ and https://github.com/raghavagps/il4pred2).

bioinformatics↗

bioRxiv2025

AdmirePred: A method for predicting abundant miRNAs in Exosomes

Non-invasive disease diagnosis is a key application of blood exosomes in liquid biopsy, as they carry diverse biological molecules, including microRNAs (miRNAs) derived from their parent cells. Developing miRNA-based disease biomarkers requires prediction of highly abundant miRNAs in exosomes under normal conditions for establishing a baseline for understanding their physiological roles and disease-specific variations. In this study, we present models for predicting highly abundant miRNAs in exosomes from their nucleotide sequences. The models were trained, tested, and evaluated on a dataset comprising 348 abundant and 349 non-abundant miRNAs. Initially, we applied alignment-based approaches, such as motif and similarity searches, but these methods yielded poor coverage. We then explored alignment-free approaches, particularly machine learning models leveraging a broad range of features. Our Extra Trees classifier, developed using binary profiles and TF-IDF features, achieved the highest performance with an AUC of 0.77. To further enhance predictive accuracy, we developed a hybrid method that combines machine learning models with alignment-based approaches, achieving an AUC of 0.854 on an independent dataset. To support research in non-invasive diagnostics and therapeutics, we have developed a web server, standalone tool, and Python package for AdmirePred, available at https://webs.iiitd.edu.in/raghava/admirepred/. Key pointsO_LImiRNA abundant in blood exosomes are promising biomarkers for liquid biopsy C_LIO_LIClassification of abundant and non-abundant miRNA in healthy individuals C_LIO_LIA hybrid method that combine alignment based and alignment free approach C_LIO_LIPrediction of miRNAs that are highly expressed in blood exosomes C_LIO_LIA web server, a python package, and a standalone tool have been created C_LI Authors BiographyO_LIAkanksha Arora is currently pursuing a Ph.D. in Computational Biology at Department of Computational Biology, Indraprastha Institute of Information Technology, New Delhi, India. C_LIO_LIGajendra P. S. Raghava is currently working as a Professor and Head of Department of Computational Biology, Indraprastha Institute of Information Technology, New Delhi, India. C_LI

bioinformatics↗

bioRxiv2025

In Silico Tool for Predicting and Scanning Rheumatoid Arthritis-Inducing Peptides in an Antigen

Rheumatoid arthritis (RA) is an autoimmune disorder in which the immune system mounts an abnormal response to self-antigens, leading to chronic inflammation and joint damage. Therefore, identifying antigenic regions in a protein that trigger RA is crucial for developing protein-based therapeutics. In this study, we developed models for predicting RA-inducing peptides using a dataset comprising of 291 experimentally confirmed RA-inducing peptides and 165 RA non-inducing peptides. Our initial analysis revealed that certain residues, such as glycine, proline, and tyrosine, are significantly enriched in RA-inducing peptides. While alignment-based techniques like BLAST and MERCI offered high precision, they suffered from limited coverage. We developed machine/deep learning based prediction and obtained highest performance (AUC = 0.75) using XGboost on an independent dataset. We also developed prediction methods using large language models and achieved highest performance (AUC 0.72) using ProtBERT. Our ensemble model achieved highest performance (AUC = 0.80 & MCC = 0.45) on an independent dataset that combine XGBoost and MERCI-derived motifs. All models were rigorously evaluated on an independent dataset not used during training or testing of models. This study will be valuable for assessing the risk of proteins used in probiotics, genetically modified foods, and protein-based therapeutics. Our most effective approach has been implemented in RAIpred, a web server and standalone software tool for predicting and scanning RA-inducing peptides. (https://webs.iiitd.edu.in/raghava/raipred/). HighlightsO_LIRheumatoid arthritis (RA), an incurable chronic joint disorder with diverse systemic complications. C_LIO_LIAn attempt to identify antigenic regions in a protein which trigger this severe disease. C_LIO_LIUtilizing sequence composition based features for developing models. C_LIO_LIImplementation of ML, DL and LLM based models for prediction of RA-inducing peptides. C_LIO_LIDevelopment of webserver, standalone, pypi and GitHub package for users. C_LI

bioinformatics↗

bioRxiv2025

A large language model for predicting pancreatic ductal adenocarcinoma patients from blood-derived exosomal transcriptomics data

Traditional machine learning approaches for text or sequence classification rely on converting textual data into numerical representations. In this study, we investigate a reverse strategy in which numerical features are transformed into sequence representations and classified using large language models (LLMs). We applied this methodology to predict pancreatic ductal adenocarcinoma (PDAC) using the expression profiles of 50 genes from 284 PDAC and 217 non-PDAC patients. Gene expression values were converted into sequence data, with each gene represented as a residue in a 50-residue protein sequence. Major LLMs like PeptideBERT, ProtBERT, and ESM2 were fine-tuned on a protein training dataset and evaluated on an independent dataset. The best-performing model, ProtBERT, achieved an AUC of 0.962 on an independent dataset. Additionally, an alignment-based approach employing BLAST and MERCI motifs was explored, and an ensemble model combining the LLM-based and alignment-based methods was developed. Our LLM-based model outperformed traditional machine learning models. To the best of our knowledge, this is the first study demonstrating the application of LLMs for mining transcriptomic profiles of cancer patients. HIGHLIGHTSO_LIIdentification of over and under-expressed genes in PDAC patients C_LIO_LIConvert numeric gene expression data to peptide sequence C_LIO_LILLM based models for predicting PDAC patients using peptide sequences C_LIO_LIMining of transcriptomics data using ProtBert and ESM2 C_LIO_LIGene expression profile for diagnostic of PDAC patients C_LI

bioinformatics↗

bioRxiv2025

NTxPred2: A large language model for predicting neurotoxic peptides and neurotoxins

Most of the existing methods including our NTxPred, use a single model to predict both neurotoxic peptides and neurotoxic proteins (neurotoxins). In this study, we developed distinct models for predicting neurotoxic peptides and neurotoxins. Our peptide dataset consists of 877 neurotoxic and 877 non-neurotoxic peptides, while our protein dataset includes 775 neurotoxins and 775 non-neurotoxins. Preliminary analysis reveals that certain residues, such as cysteine, are more prevalent in both neurotoxic peptides and proteins, though their abundance differs in magnitude. First, we developed machine learning models using composition and binary profiles, achieving a maximum AUC of 0.97 for peptides and 0.85 for proteins. The performance for proteins improved from an AUC of 0.85 to 0.89 when evolutionary information was incorporated. Next, we built machine learning models using embeddings from protein language models, attaining an AUC of 0.96 for peptides and 0.94 for proteins. We also developed protein language models and achieved an AUC of 0.98 for peptides using esm2-t30 and 0.91 for proteins using esm2-t6. All models were trained and tested using 5-fold cross-validation, and final models were evaluated on an independent dataset not used in training. We further assessed protein models on the peptide dataset and vice versa, highlighting the necessity of separate models. The proposed models outperform existing methods on independent datasets. Our neurotoxicity prediction models will aid in the safety assessment of genetically modified foods and therapeutic proteins by minimizing the need for animal testing. To support the scientific community, we developed a standalone software and web server NTxPred2, for predicting and scanning neurotoxins (https://webs.iiitd.edu.in/raghava/ntxpred2/, https://github.com/raghavagps/ntxpred2/). HighlightsO_LIDevelopment of models for predicting neurotoxic peptides and neurotoxins C_LIO_LIML Models Using LLM Embeddings for Neurotoxin Prediction C_LIO_LIProtein language models esm2-t30 for predicting neurotoxic peptides C_LIO_LIImportant for risk assessment of genetically modified foods/proteins C_LIO_LINTxPred2 is an updated version of NTxPred for predicting neurotoxicity C_LI

bioinformatics↗

bioRxiv2025

Identification of Multiple Prognostic Biomarker Sets for Risk Stratification in SKCM

Most of the existing studies have identified a single profile of prognostic biomarkers for predicting high-risk cancer patients using transcriptomics data. In this study, we propose multiple distinct sets of prognostic biomarkers for predicting high-risk Skin Cutaneous Melanoma (SKCM) patients. Our primary analysis reveals that the expression of certain genes, such as CREG1, PCGF5, and VPS13C, strongly correlates with overall survival (OS) in SKCM patients. We developed machine learning-based prognostic models to predict 1-, 3-, and 5-year overall survival using gene expression profiles. State-of-the-art feature selection techniques were employed to identify the first prognostic biomarkers set consisting of 20 genes. Machine learning models built to predict high-risk patients using this set of biomarkers and achieved an AUC of 0.90 and a Kappa of 0.58 in distinguishing high-risk SKCM patients. Similarly, a second independent set of 20 prognostic genes was identified, with no overlap with the first set. The best model trained on this second set achieved an AUC of 0.89 with a Kappa of 0.56, while the fifth biomarker set achieved the highest performance with an AUC of 0.91 and a Kappa of 0.64. This process was repeated to obtain a total of seven distinct prognostic biomarker sets, each containing 20 unique genes. The predictive performance of these models varied between 0.84 and 0.91 for AUC and 0.48 to 0.64 for Kappa on the validation dataset. These findings demonstrate that it is possible to identify multiple independent sets of prognostic biomarkers, each capable of accurately predicting high-risk SKCM patients. HighlightsO_LIPrediction of high-risk SKCM patients from transcriptomics data C_LIO_LIIdentification of multiple set of prognostic biomarkers C_LIO_LIIdentification of set of biomarkers using feature selection technique C_LIO_LIDevelopment of machine learning based models for prediction C_LIO_LIPathway analysis of selected set of biomarker genes C_LI

bioinformatics↗

bioRxiv2025

NfκBin: A Machine Learning based Method for Screening Nf-κB Inhibitors.

Nuclear Factor kappa B (NF-{kappa}B) is a transcription factor whose upregulation is associated in chronic inflammatory diseases, including rheumatoid arthritis, inflammatory bowel disease, and asthma. In order to develop therapeutic strategies targeting NF-{kappa}B-related diseases, we developed a computational approach to predict drugs capable of inhibiting NF-{kappa}B signaling pathways. In this study, we utilized a dataset comprising 1,149 inhibitors and 1,332 non-inhibitors retrieved from PubChem. Chemical descriptors were computed using the PaDEL software, and relevant features were selected using advanced feature selection techniques. Initially, machine learning models were constructed using 2D descriptors, 3D descriptors, and molecular fingerprints, achieving maximum AUC values of 0.66, 0.56, and 0.66, respectively. To improve feature selection, we applied univariate analysis and SVC-L1 regularization to identify features that can effectively differentiate inhibitors from non-inhibitors. Using these selected features, we developed machine learning models, our support vector classifier achieved an AUC of 0.75 on the validation dataset. All models were trained using five-fold cross-validation and rigorously evaluated on an independent validation dataset, which was not used during training or hyperparameter optimization. Finally, our best-performing model was employed to screen FDA-approved drugs for potential NF-{kappa}B inhibitors. Notably, most of the predicted inhibitors corresponded to drugs previously identified as inhibitors in experimental studies, underscoring the models predictive reliability. Our best- performing models have been integrated into a standalone software and web server, Nf{kappa}Bin, to enable the scientific community to perform high-throughput screening of Nf-{kappa}B inhibitors from chemical libraries. (https://webs.iiitd.edu.in/raghava/nfkbin/). Highlights1) Inhibitors of Nf-{kappa}B can be used to manage disease like rheumatoid arthritis, asthma. 2) A novel tool Nf{kappa}Bin developed for screening Nf-{kappa}B inhibitors from chemical library. 3) Cutting-edge feature selection techniques used to identify the most relevant chemical descriptors. 4) Models were developed and validated on a robust dataset of inhibitors and non-inhibitors. 5) Integrated best-performing models into the Nf{kappa}Bin software and web server.

bioinformatics↗

bioRxiv2024

HAIRpred: Prediction of human antibody interacting residues in an antigen from its primary structure

1.In the past, several methods have been developed for predicting conformational B-cell epitopes in antigens that are not specific to any host. Our primary analysis of antibody-antigen complexes indicated a need to develop host-specific B-cell epitopes. In this study, we present a novel approach to predict conformational B-cell epitopes specific to human hosts by focusing on human antibody-interacting residues in antigens. We trained, test and evaluate our models on 277 complexes of human antibody-antigen complexes. Initially, we employed machine learning models based on the binary sequence profile of antigens, achieving a maximum area under the receiver operating characteristic curve (AUC) of 0.61. Performance of model improved significantly AUC from 0.61 to 0.67, when evolutionary profiles are used instead of binary profiles. Models developed using embeddings from fine-tuned large language models reached an AUC of 0.61. Additionally, models utilizing predicted surface relative solvent accessibility achieved an AUC of 0.67. Our ensemble model, which combined relative surface accessibility with evolutionary profiles, achieved highest precision with an AUCROC of 0.72. All models in this study were trained using five-fold cross-validation on a training dataset and evaluated on an independent dataset not used for training or testing. Our method outperforms existing approaches on the independent dataset. Furthermore, we used SHAP eXplainable AI (XAI) method to interpret the importance of individual features contributing to the predictions made by our models. To support the scientific community, we have developed a standalone software, and web server, HAIRpred, for predicting human antibody-interacting residues in proteins. https://webs.iiitd.edu.in/raghava/hairpred/. Authors BiographyO_LIRuchir Sahni is currently studying as an integrated BS-MS student at Indian Institute of Science Education and Research (IISER) Pune, India. He is currently working as an Intern on Project position at Department of Computational Biology, Indraprastha Institute of Information Technology (IIIT), New Delhi, India. C_LIO_LINishant Kumar is currently working as Ph.D. in Computational biology from Department of Computational Biology, Indraprastha Institute of Information Technology (IIIT), New Delhi, India. C_LIO_LIGajendra P. S. Raghava is currently working as Professor and Head of Department of Computational Biology, Indraprastha Institute of Information Technology (IIIT), New Delhi, India. C_LI

bioinformatics↗

bioRxiv2024

An ensemble method for predicting and designing of druggable proteins.

In the past, numerous proteins/peptides have been discovered, which have a wide range of therapeutic properties like anticancer, antimicrobial, antihypertensive. Only few hundreds of proteins are druggable (approved by US FDA), most of the proteins fails in clinical trials. In this study, an attempt had been made to understand properties of FDA approved proteins to develop models for predicting druggable proteins. Our main dataset 356 FDA approved proteins as positive dataset and equal number of randomly selected proteins as negative dataset. We used 80% data for training and 20% for independent validation, no protein in validation dataset have more than 40% similarity with any protein in training dataset. We deployed machine learning based models using a five-fold cross-validation and test on validation dataset. Our random forest-based model developed using SVC-L1 selected features obtained maximum performance AUC of 0.80 with MCC 0.61 on validation data. In addition to this, we performed MERCI-based motif analysis to find motifs in druggable proteins. Finally, we developed an ensemble-based method combining best performing machine learning model with motifs and achieved AUC 0.92 with MCC 0.83 on independent validation dataset. We developed a web server and standalone package ThPPred to facilitate scientific community in predicting and designing druggable proteins (https://webs.iiitd.edu.in/raghava/thppred/). HighlightsO_LIAnalysis of FDA approved or druggable proteins C_LIO_LIDiscrimination of druggable and non-druggable proteins C_LIO_LIMachine learning based models for predicting druggable molecules C_LIO_LIIdentification of motifs in druggable proteins C_LIO_LIA web server for providing service to community C_LI

bioinformatics↗

bioRxiv2024

CytoLNCpred - A computational method for predicting cytoplasm associated long-coding RNAs in 15 cell-lines

The function of long non-coding RNA (lncRNA) is largely determined by its specific location within a cell. Previous methods have used noisy datasets, including mRNA transcripts in tools intended for lncRNAs, and excluded lncRNAs lacking significant differential localization between the cytoplasm and nucleus. In order to overcome these shortcomings, a method has been developed for predicting cytoplasm-associated lncRNAs in 15 human cell-lines, identifying which lncRNAs are more abundant in the cytoplasm compared to the nucleus. All models in this study were trained using five-fold cross validation and tested on an independent dataset. Initially, we developed machine and deep learning based models using traditional features like composition and correlation. Using composition and correlation based features, machine learning algorithms achieved an average AUC of 0.7049 and 0.7089, respectively for 15 cell-lines. Secondly, we developed machine based models developed using embedding features obtained from the large language model DNABERT-2. The average AUC for all the cell-lines achieved by this approach was 0.6604. Subsequently, we also fine-tuned DNABERT-2 on our training dataset and evaluated the fine-tuned DNABERT-2 model on the independent dataset. The fine-tuned DNABERT-2 model achieved an average AUC of 0.6336. Correlation-based features combined with ML algorithms outperform LLM-based models, in the case of predicting differential lncRNA localization. These cell-line specific models as well as web-based service are available to the public from our web server (https://webs.iiitd.edu.in/raghava/cytolncpred/) . HIGHLIGHTSO_LIPrediction of cytoplasm-associated lncRNAs in 15 human cell lines C_LIO_LIMachine learning using composition and correlation features C_LIO_LIDNABERT-2 embeddings for lncRNA localization prediction C_LIO_LICorrelation-based models outperform LLM-based models C_LIO_LIWeb server and models available for public use C_LI AUTHORS BIOGRAPHYO_LIShubham Choudhury is currently working as Ph.D. in Computational Biology from Department of Computational Biology, Indraprastha Institute of Information Technology, New Delhi, India. C_LIO_LINaman Kumar Mehta is currently working as Ph.D. in Computational Biology from Department of Computational Biology, Indraprastha Institute of Information Technology, New Delhi, India. C_LIO_LIGajendra P. S. Raghava is currently working as Professor and Head of Department of Computational Biology, Indraprastha Institute of Information Technology, New Delhi, India C_LI

bioinformatics↗

bioRxiv2024

Designing of thermostable proteins with a desired melting temperature

1.The stability of proteins at higher temperatures is crucial for its functionality that is measured by their melting temperature (Tm). The Tm is the temperature at which 50% of the protein loses its native structure and activity. Existing methods for predicting Tm have two major limitations: first, they are often trained on redundant proteins, and second, they do not allow users to design proteins with the desired Tm. To address these limitations, we developed a regression method for predicting the Tm value of proteins using 17,312 non-redundant proteins, where no two proteins are more than 40% similar. We used 80% of the data for training and testing; remaining 20% of the data for validation. Initially, we developed a machine learning model using standard features from protein sequences. Our best model, developed using Shannon entropy for all residues, achieved the highest Pearson correlation of 0.80 with an R{superscript 2} of 0.63 between the predicted and actual Tm of proteins on the validation dataset. Next, we fine-tuned large language models (e.g., ProtBert, ProtGPT2, ProtT5) on our training dataset and generated embeddings. These embeddings have been used for developing machine learning models. Our best model, developed using ProtBert embeddings, achieved a maximum correlation of 0.89 with an R{superscript 2} of 0.80 on the validation dataset. Finally, we developed an ensemble method that combines standard protein features and embeddings. One of the aims of the study is to assist the scientific community in the design of targeted melting temperatures. We created a user-friendly web server and a python package for predicting and designing thermostable proteins. Our standalone software can be used to screen thermostable proteins in genomes and metagenomes. We demonstrated the application of PPTstab in identifying thermostable proteins in different organisms from their genomes, the model and data is available at: https://webs.iiitd.edu.in/raghava/pptstab. HighlightsO_LIPrediction of melting temperature (Tm) on non-redundant proteins C_LIO_LIMachine learning models based on sequence composition and ProtBert embeddings C_LIO_LIA Webserver for predicting Tm and designing thermostable proteins C_LI

bioinformatics↗

bioRxiv2024

Prediction of Hemolytic Peptides and their Hemolytic Concentration (HC50)

Several peptide-based drugs fail in clinical trials due to their toxicity or hemolytic activity against red blood cells (RBCs). Existing methods predict hemolytic peptides but not the concentration (HC50) required to lyse 50% of RBCs. In this study, we developed a classification model and regression model to identify and quantify the hemolytic activity of peptides. Our models were trained and validated on 1924 peptides with experimentally determined HC50 against mammalian RBCs. Analysis indicates that hydrophobic and positively charged residues were associated with higher hemolytic activity. Our classification models achieved a maximum AUC of 0.909 using a hybrid model of ESM-2 and a motif-based approach. Regression models using compositional features achieved R of 0.739 with R{superscript 2} of 0.543. Our models outperform existing methods and are implemented in the web-based platform HemoPI2 and standalone software for designing hemolytic peptides with desired HC50 values (http://webs.iiitd.edu.in/raghava/hemopi2/). HighlightsO_LIDeveloped classification and regression models to predict hemolytic activity and HC50 values of peptides. C_LIO_LIA hybrid model combining machine learning and motif prediction excels in accuracy. C_LIO_LIBenchmarking of the existing classification methods on independent datasets. C_LIO_LIWeb server, standalone software, and pip package for hemolytic activity prediction of peptides/proteins. C_LI

bioinformatics↗

bioRxiv2024

Prediction of plant resistance proteins using alignment-based and alignment-free approaches

Plant Disease Resistance (PDR) proteins are critical in identifying and killing plant pathogens. Predicting PDR protein is essential for understanding plant-pathogen interactions and developing strategies for crop protection. This study proposes a hybrid model for predicting and designing PDR proteins against plant-invading pathogens. Initially, we tried alignment-based approaches, such as BLAST for similarity search and MERCI for motif search. These alignment-based approaches exhibit very poor coverage or sensitivity. To overcome these limitations, we developed alignment-free or machine learning-based methods using compositional features of proteins. Our machine learning-based model, developed using compositional features of proteins, achieved a maximum performance AUROC of 0.92. The performance of our model improved significantly from AUROC of 0.92 to 0.95 when we used evolutionary information instead of protein sequence. Finally, we developed a hybrid or ensemble model that combined our best machine learning model with BLAST and obtained the highest AUROC of 0.98 on the validation dataset. We trained and tested our models on a training dataset and evaluated them on a validation dataset. None of the proteins in our validation dataset are more than 40% similar to proteins in the training dataset. One of the objectives of this study is to facilitate the scientific community working in plant biology. Thus, we developed an online platform for predicting and designing plant resistance proteins, "PlantDRPpred" (https://webs.iiitd.edu.in/raghava/plantdrppred). HighlightsO_LIDevelopment of a Machine-learning model for resistance protein prediction. C_LIO_LIUsed alignment-based and alignment-free ensemble methods. C_LIO_LIWeb server development and standalone package. C_LIO_LIPrediction and design of PDR proteins. C_LI

bioinformatics↗

bioRxiv2024

Prediction of inhibitory peptides against E. coli with desired MIC value

In the past, several methods have been developed for predicting antibacterial and antimicrobial peptides, but only limited attempts have been made to predict their minimum inhibitory concentration (MIC) values. In this study, we trained our models on 3,143 peptides and validated them on 786 peptides whose MIC values have been determined experimentally against Escherichia coli (E. coli). The correlational analysis reveals that the Composition Enhanced Transition and Distribution (CeTD) attributes strongly correlate with MIC values. We initially employed the similarity search strategy utilizing BLAST to estimate MIC values of peptides but found it inadequate for prediction. Next, we developed machine learning techniques-based regression models using a wide range of features, including peptide composition, binary profile, and embeddings of large language models. We implemented feature selection techniques like minimum Redundancy Maximum Relevance (mRMR) to select the best relevant features for developing prediction models. Our Random forest-based regressor, based on selected features, achieved a correlation coefficient (R) of 0.78, R-squared (R{superscript 2}) of 0.59, and a root mean squared error (RMSE) of 0.53 on the validation dataset. Our best model outperforms the existing methods when benchmarked on an independent dataset of 498 inhibitory peptides of E. coli. One of the major features of the web-based platform EIPpred developed in this study is that it allows users to identify or design peptides that can inhibit E. coli with the desired MIC value (https://webs.iiitd.edu.in/raghava/eippred). HighlightsO_LIPrediction of MIC value of peptides against E.coli. C_LIO_LIAn independent dataset was generated for comparison. C_LIO_LIFeature selection using the mRMR method. C_LIO_LIA regressor method for designing novel inhibitory peptides. C_LIO_LIA web server and standalone package for predicting the inhibitory activity of peptides. C_LI

bioinformatics↗

bioRxiv2024

Prediction of exosomal miRNA-based biomarkers for liquid biopsy

In this study, we investigated the properties of exosomal miRNAs to identify potential biomarkers for liquid biopsy. We collected 956 exosomal and 956 non-exosomal miRNA sequences from RNALocate and miRBase to develop predictive models. Our initial analysis reveals that specific nucleotides are preferred at certain positions in miRNAs associated with exosomes. We employed an alignment-based approach, artificial intelligence (AI) models, and ensemble methods for predicting exosomal miRNAs. For the alignment-based approach, we used a motif-based method with MERCI and a similarity-based method with BLAST, achieving high precision but low coverage of about 29%. The AI models, developed using machine learning, deep learning techniques, and large language models, achieved a maximum AUC of 0.707 and an MCC of 0.268 on an independent dataset. Finally, our ensemble method, combining alignment-based and AI-based models, reached a maximum AUC of 0.73 and an MCC of 0.352 on an independent dataset. We have developed a web server, EmiRPred, to assist the scientific community in predicting and designing exosomal miRNAs and identifying associated motifs (https://webs.iiitd.edu.in/raghava/emirpred/). Key pointsO_LIExosomal miRNAs have potential applications in liquid biopsy C_LIO_LIAn ensemble method has been developed to predict and design exosomal miRNA C_LIO_LIAn array of predictive models were built using alignment-based approaches and AI-based approaches (ML, DL, LLM) C_LIO_LIA variety of important features and motifs for exosomal miRNA have been identified C_LIO_LIA webserver, a python package, a github, and a standalone software have been created C_LI

bioinformatics↗

bioRxiv2024

Prediction of anti-freezing proteins from their evolutionary profile

1.Prediction of antifreeze proteins (AFPs) holds significant importance due to their diverse applications in healthcare. An inherent limitation of current AFP prediction methods is their reliance on unreviewed proteins for evaluation. This study evaluates proposed and existing methods on an independent dataset containing 81 AFPs and 73 non-AFPs obtained from Uniport, which have been already reviewed by experts. Initially, we constructed machine learning models for AFP prediction using selected composition-based protein features and achieved a peak AUC of 0.90 with an MCC of 0.69 on the independent dataset. Subsequently, we observed a notable enhancement in model performance, with the AUC increasing from 0.90 to 0.93 upon incorporating evolutionary information instead of relying solely on the primary sequence of proteins. Furthermore, we explored hybrid models integrating our machine learning approaches with BLAST-based similarity and motif-based methods. However, the performance of these hybrid models either matched or was inferior to that of our best machine-learning model. Our best model based on evolutionary information outperforms all existing methods on independent/validation dataset. To facilitate users, a user-friendly web server with a standalone package named "AFPropred" was developed (https://webs.iiitd.edu.in/raghava/afpropred). HighlightsO_LIPrediction of antifreeze proteins with high precision C_LIO_LIEvaluation of prediction models on an independent dataset C_LIO_LIMachine learning based models using sequence composition C_LIO_LIEvolutionary information based prediction models C_LIO_LIA webserver for predicting, scanning, and designing AFPs. C_LI Authors BiographyO_LINishant Kumar is currently working as Ph.D. in Computational biology from Department of Computational Biology, Indraprastha Institute of Information Technology, New Delhi, India. C_LIO_LIShubham Choudhury is currently working as Ph.D. in Computational biology from Department of Computational Biology, Indraprastha Institute of Information Technology, New Delhi, India C_LIO_LINisha Bajiya is currently working as Ph.D. in Computational biology from Department of Computational Biology, Indraprastha Institute of Information Technology, New Delhi, India C_LIO_LISumeet Patiyal is currently working as a postdoctoral visiting fellow Cancer Data Science Laboratory, National Cancer Institute, National Institutes of Health, Bethesda, Maryland, USA. C_LIO_LIGajendra P. S. Raghava is currently working as Professor and Head of Department of Computational Biology, Indraprastha Institute of Information Technology, New Delhi, India. C_LI

bioinformatics↗

bioRxiv2023

Molecular Property Diagnostic Suite for COVID-19 (MPDSCOVID-19): An open access disease specific drug discovery portal

Computational drug discovery is intrinsically interdisciplinary and has to deal with the multifarious factors which are often dependent on the type of disease. Molecular Property Diagnostic Suite (MPDS) is a Galaxy based web portal which was conceived and developed as a disease specific web portal, originally developed for tuberculosis (MPDSTB). As specific computational tools are often required for a given disease, developing a disease specific web portal is highly desirable. This paper emphasises on the development of the customised web portal for COVID-19 infection and is referred to as MPDSCOVID-19. Expectedly, the MPDS suites of programs have modules which are essentially independent of a given disease, whereas some modules are specific to a particular disease. In the MPDSCOVID-19 portal, there are modules which are specific to COVID-19, and these are clubbed in SARS-COV-2 disease library. Further, the new additions and/or significant improvements were made to the disease independent modules, besides the addition of tools from galaxy toolshed. This manuscript provides a latest update on the disease independent modules of MPDS after almost 6 years, as well as provide the contemporary information and tool-shed necessary to engage in the drug discovery research of COVID-19. The disease independent modules include file format converter and descriptor calculation under the data processing module; QSAR, pharmacophore, scaffold analysis, active site analysis, docking, screening, drug repurposing tool, virtual screening, visualisation, sequence alignment, phylogenetic analysis under the data analysis module; and various machine learning packages, algorithms and in-house developed machine learning antiviral prediction model are available. The MPDS suite of programs are expected to bring a paradigm shift in computational drug discovery, especially in the academic community, guided through a transparent and open innovation approach. The MPDSCOVID-19 can be accessed at http://mpds.neist.res.in:8085.

bioinformatics↗