CLMB: Deep Contrastive Learning for Robust Metagenomic Binning

Zhang, Pengfei; Jiang, Zhengyuan; Wang, Yixuan; Yu, Li

doi:10.1007/978-3-031-04749-7_23

Cited by 8 publications

(11 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Based on the VAMB architecture, various methods have been developed for its extension or the use of other sources of information. First, the authors of CLMB [ 114 ] took into account the noise, rarely considered in metagenomic analysis. To do so, they simulated different types of noise, augmenting contig data with noised sequences.…”

Section: Resultsmentioning

confidence: 99%

Deep learning methods in metagenomics: a review

Roy,

Prifti,

Belda

et al. 2024

Microbial Genomics

View full text Add to dashboard Cite

The ever-decreasing cost of sequencing and the growing potential applications of metagenomics have led to an unprecedented surge in data generation. One of the most prevalent applications of metagenomics is the study of microbial environments, such as the human gut. The gut microbiome plays a crucial role in human health, providing vital information for patient diagnosis and prognosis. However, analysing metagenomic data remains challenging due to several factors, including reference catalogues, sparsity and compositionality. Deep learning (DL) enables novel and promising approaches that complement state-of-the-art microbiome pipelines. DL-based methods can address almost all aspects of microbiome analysis, including novel pathogen detection, sequence classification, patient stratification and disease prediction. Beyond generating predictive models, a key aspect of these methods is also their interpretability. This article reviews DL approaches in metagenomics, including convolutional networks, autoencoders and attention-based models. These methods aggregate contextualized data and pave the way for improved patient care and a better understanding of the microbiome’s key role in our health.

show abstract

Section: Resultsmentioning

confidence: 99%

Deep learning methods in metagenomics: a review

Roy,

Prifti,

Belda

et al. 2024

Microbial Genomics

View full text Add to dashboard Cite

show abstract

“…CL has been effectively applied in various domains such as self-supervised learning [Chen et al, 2020, Chuang et al, 2020], image classification [Khosla et al, 2020], gaze direction regression [Wang et al, 2022] and metagenomics [Zhang et al, 2022]. In this work, we adapt CL to compare different RNA-seq samples.…”

Section: Related Workmentioning

confidence: 99%

Data-driven AI system for learning how to run transcript assemblers

Shen,

Yan,

Kingsford

2024

Preprint

View full text Add to dashboard Cite

Transcript assemblers are tools to reconstruct expressed transcripts from RNA-seq data. These tools have a large number of tunable parameters, and accurate transcript assembly requires setting them suitably. Because of the heterogeneity of different RNA-seq samples, a single default setting or a small fixed set of parameter candidates can only support the good performance of transcript assembly on average, but are often suboptimal for many individual samples. Manually tuning parameters for each sample is time consuming and requires specialized experience. Therefore, developing an automated system that can advise good parameter settings for individual samples becomes an important problem. Using Bayesian optimization and contrastive learning, we develop a new automated parameter advising system for transcript assembly that can generate sets of sample-specific good parameter candidates. Our framework achieves efficient sample-specific parameter advising by learning parameter knowledge from a representative set of existing RNA-seq samples and transferring the knowledge to unseen samples. We use Scallop and StringTie, two well-known transcript assemblers, to test our framework on two collections of RNA-seq samples. Results show that our new parameter advising system significantly outperforms the previous advising method in each dataset and each transcript assembler. Scripts to reproduce the results from this study can be found at https://github.com/Kingsford-Group/autoparadvisor.

show abstract

“…The proposed model employs convolutional layers with rectified linear unit (ReLU) activation functions, in addition to incorporating max-pooling layers inside the encoder portion. This choice of a convolutional autoencoder is made to improve computational complexity and overall performance [37]. The encoder component of the model employs a non-linear transformation to convert the input vector into a lower-dimensional hidden representation.…”

Section: Stacked Convolutional Autoencoder (Scae)mentioning

confidence: 99%

Enhancing Taxonomic Categorization of DNA Sequences with Deep Learning: A Multi-Label Approach

Hossain,

Kim,

Uddin

et al. 2023

Bioengineering

View full text Add to dashboard Cite

The application of deep learning for taxonomic categorization of DNA sequences is investigated in this study. Two deep learning architectures, namely the Stacked Convolutional Autoencoder (SCAE) with Multilabel Extreme Learning Machine (MLELM) and the Variational Convolutional Autoencoder (VCAE) with MLELM, have been proposed. These designs provide precise feature maps for individual and inter-label interactions within DNA sequences, capturing their spatial and temporal properties. The collected features are subsequently fed into MLELM networks, which yield soft classification scores and hard labels. The proposed algorithms underwent thorough training and testing on unsupervised data, whereby one or more labels were concurrently taken into account. The introduction of the clade label resulted in improved accuracy for both models compared to the class or genus labels, probably owing to the occurrence of large clusters of similar nucleotides inside a DNA strand. In all circumstances, the VCAE-MLELM model consistently outperformed the SCAE-MLELM model. The best accuracy attained by the VCAE-MLELM model when the clade and family labels were combined was 94%. However, accuracy ratings for single-label categorization using either approach were less than 65%. The approach’s effectiveness is based on MLELM networks, which record connected patterns across classes for accurate label categorization. This study advances deep learning in biological taxonomy by emphasizing the significance of combining numerous labels for increased classification accuracy.

show abstract

CLMB: Deep Contrastive Learning for Robust Metagenomic Binning

Cited by 8 publications

References 38 publications

Deep learning methods in metagenomics: a review

Deep learning methods in metagenomics: a review

Data-driven AI system for learning how to run transcript assemblers

Enhancing Taxonomic Categorization of DNA Sequences with Deep Learning: A Multi-Label Approach

Contact Info

Product

Resources

About