Article
Computer Science, Artificial Intelligence
Hongpeng Yang, Yijie Ding, Jijun Tang, Fei Guo
Summary: In this study, a model based on Multiple Kernel fusion on Graph Convolutional Network (MKGCN) is proposed for inferring novel microbe-drug associations. The model extracts multi-layer features from the heterogeneous network of microbes and drugs using Graph Convolutional Network (GCN), calculates kernel matrices on each layer, and fuses them through an average weighting method. The model shows excellent prediction performance on three datasets and is validated through a case study on the SARS-Cov-2 virus.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Biochemistry & Molecular Biology
Yuttapong Thawornwattana, Jun Huang, Tomas Flouri, James Mallet, Ziheng Yang
Summary: Genomic data provide valuable information for studying species divergence and gene flow between species, including the direction, timing, and strength of gene flow. However, inferring the direction of gene flow is challenging due to similar patterns generated by gene flow in opposite directions. This study uses likelihood-based methods and simulation to investigate the information about the direction of gene flow present in genomic sequence data. The results show that it is easier to infer gene flow from a small population to a large one and from outgroup species to ingroup species than in the opposite direction, and a longer time of separate evolution between divergence and introgression makes gene flow inference easier.
MOLECULAR BIOLOGY AND EVOLUTION
(2023)
Article
Biochemical Research Methods
Lizhi Liu, Hiroshi Mamitsuka, Shanfeng Zhu
Summary: Deciphering the relationship between human genes/proteins and abnormal phenotypes is crucial for disease prevention and treatment, requiring computational predictions. The HPODNets model, with features including multiple network input, semi-supervised learning, and deep graph convolutional network, outperforms other methods in predicting human protein-phenotype associations.
Article
Mathematical & Computational Biology
Lei Chen, Kaiyu Chen, Bo Zhou
Summary: Drugs are essential for human health, but developing new drugs is laborious and expensive. Drug repositioning, which discovers novel effects of existing drugs, is seen as an effective way to accelerate drug development. This study introduces a novel reliable negative sample selection scheme (RNSS) that selects pairs of drugs and diseases with low probabilities of being actual drug-disease associations. The cross-validation results demonstrate that classifiers built using negative samples selected by RNSS perform nearly perfectly and outperform traditional and previous selection schemes.
MATHEMATICAL BIOSCIENCES AND ENGINEERING
(2023)
Article
Multidisciplinary Sciences
Madikay Senghore, Hannah Read, Priyali Oza, Sarah Johnson, Hemanoel Passarelli-Araujo, Bradford P. Taylor, Stephen Ashley, Alex Grey, Alanna Callendrello, Robyn Lee, Matthew R. Goddard, Thomas Lumley, William P. Hanage, Siouxsie Wiles
Summary: Identifying and interrupting transmission chains is crucial for controlling infectious diseases. This study assesses the use of deep-sequenced genomic surveillance to identify transmission pairs and highlights the importance of within-host variation in transmission.
NATURE COMMUNICATIONS
(2023)
Article
Genetics & Heredity
Alec M. Chiu, Erin K. Molloy, Zilong Tan, Ameet Talwalkar, Sriram Sankararaman
Summary: Inferring the structure of human populations from genetic variation data is crucial in population and medical genomic research. A new method called SCOPE is introduced, which is significantly faster than existing methods while maintaining comparable accuracy. SCOPE can infer population structure within a day for datasets containing millions of individuals and genetic variants, and it can also utilize allele frequencies from previous studies to improve the interpretability of the results.
AMERICAN JOURNAL OF HUMAN GENETICS
(2022)
Article
Biotechnology & Applied Microbiology
Abdou Rahmane Wade, Harold Durufle, Leopoldo Sanchez, Vincent Segura
Summary: Using 241 poplar genotypes, this study generated large datasets of phenotypic, genomic, and transcriptomic information, and built prediction models for traits based on SNPs and transcripts. Integration analysis revealed a negative correlation between changes in predictability and predictor ranking for trans eQTLs. This study provides a novel method to explore data integration.
Article
Biochemical Research Methods
Minwoo Pak, Sangseon Lee, Inyoung Sung, Bonil Koo, Sun Kim
Summary: Drug response prediction (DRP) is crucial for precision medicine to anticipate patient reactions to drugs. While most studies use cell line transcriptome data and drug chemical structures to predict drug response, this study proposes a framework that leverages drug target interaction (DTI) information to improve deep learning-based DRP models. By computing gene perturbation scores through network propagation techniques, the framework integrates this DTI information with existing DRP models. The results show significant performance boosts, especially when dealing with unknown drugs.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Biology
Xiuhong Li, Hao Yuan, Xiaoliang Wu, Chengyi Wang, Hongbo Shi, Yingli Lv
Summary: Metabolic processes in the human body are crucial for maintaining normal life activities, and abnormal concentrations of metabolites are closely linked to disease occurrence and development. Drug usage has a significant impact on metabolism, as drug metabolites can influence efficacy, toxicity, and interactions. However, our understanding of metabolite-drug associations remains incomplete, and individual data sources are often incomplete and noisy. Thus, there is an urgent need to integrate various data sources to infer reliable metabolite-drug associations.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Multidisciplinary Sciences
Nicolas Gonzalez-Arrue, Isidora Inostroza, Raul Conejeros, Marcelo Rivas-Astroza
Summary: A cell's genome influences its metabolism, but there are post-transcriptional mechanisms that also regulate reaction kinetics. We presented Pheflux, a model that accounts for unobserved mechanisms of reaction kinetics when inferring the fluxome from a transcriptome. Pheflux accurately estimates the carbon core metabolism and shows higher glucose yields on lactate in tumoral cells compared to normal counterparts, consistent with the Warburg effect.
Review
Biochemical Research Methods
Le Ou-Yang, Dehan Cai, Xiao-Fei Zhang, Hong Yan
Summary: The mechanisms controlling biological processes can be investigated through the changes in gene dependency networks. The novel weighted differential network estimation (WDNE) model shows superior performance in handling gene expression data with missing values and changes in expression levels.
BRIEFINGS IN BIOINFORMATICS
(2021)
Article
Biochemistry & Molecular Biology
Yizhan Li, Runqi Wang, Shuo Zhang, Hanlin Xu, Lei Deng
Summary: Accurate inference of the relationship between non-coding RNAs and drug resistance is crucial for understanding complex mechanisms of drug actions. In this paper, a novel computational method LRGCPND is proposed, which outperforms seven other state-of-the-art approaches with an average AUC value of 0.8987. Results show that LRGCPND is an effective tool for inferring associations between ncRNA and drug resistance.
INTERNATIONAL JOURNAL OF MOLECULAR SCIENCES
(2021)
Article
Chemistry, Multidisciplinary
Satanat Kitsiranuwat, Apichat Suratanee, Kitiporn Plaimas
Summary: This study proposed a method for drug repositioning based on protein-protein similarity vectors (PPSVs) and classification techniques, achieving high performance in predicting disease characteristics and identifying new potential drugs. The interaction confidence value in the protein-protein interaction network and functional pathway score were found to be key descriptors for prediction in the functional aspect.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Information Systems
Eunji Jeon, So Young Sohn
Summary: Digital therapeutics companies are using a link prediction framework to reposition their products and identify new target diseases. By integrating multiple data sources and applying graph embedding methods, new indications for DTx products can be suggested based on latent features of disorders. This study applied the framework to psychiatric DTx products, determining new target disorders with high treatment potential. The results are expected to assist DTx firms in entering novel markets and improve overall patient access and public health.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Biochemistry & Molecular Biology
Alisa Pavel, Laura A. Saarimaki, Lena Mobus, Antonio Federico, Angela Serra, Dario Greco
Summary: The analysis of large integrated data sets in life sciences faces challenges, and the integration of various research angles and data types is crucial, especially in drug development and chemical safety assessment where computational methods can provide solutions.
COMPUTATIONAL AND STRUCTURAL BIOTECHNOLOGY JOURNAL
(2022)