We note that forward and reverse-complement representations can be merged by any element-wise operation on the 1D output tensors. Similarly to Shrikumar et al. (2017), we place the dense layers (and the output layer) after representation merging. Apart from summation, we consider two alternative merging functions (note that averaging and adding the representation vectors are essentially equivalent). The max function implements the Gödel t-conorm, corresponding to the OR operation in Gödel fuzzy logic. Even though the activations are not restricted to the interval [0,1], high output values can be interpreted as finding a motif on either of the two strands. The Hadamard product is the product t-norm corresponding to the AND operation in the product fuzzy logic. Here, high values may be understood as finding a motif on both strands at the same time.2 Training and tuning 2.1 Class weighting While the data preprocessing procedure results in a balanced training dataset, the mean coverage of pathogen and non-pathogen genomes is drastically different. To explore an alternative way of solving the class imbalance problem, we also simulated an imbalanced training set, where the total of 20 million training reads was simulated with equal mean coverage from all the training genomes, regardless of their labels. This dataset was then used to train ten networks using a class-weighted loss function. PaPrBaG, constituting the state-of-the art in machine learning based pathogenicity prediction, does not support error weighting. For BLAST, a method based on sequence homology, this distinction does not apply at all, as its reference database is constructed over whole genomes. However, this does not influence our final results; the class-weighted networks 1

AbstractViruses evolve extremely quickly, so reliable methods for viral host prediction are necessary to safeguard biosecurity and biosafety alike. Novel human-infecting viruses are difficult to detect with standard bioinformatics workflows. Here, we predict whether a virus can infect humans directly from next-generation sequencing reads. We show that deep neural architectures significantly outperform both shallow machine learning and standard, homology-based algorithms, cutting the error rates in half and generalizing to taxonomic units distant from those presented during training. We propose a new approach for convolutional filter visualization to disentangle the information content of each nucleotide from its contribution to the final classification decision. Nucleotide-resolution maps of the learned associations between pathogen genomes and the infectious phenotype can be used to detect virulence-related genes in novel agents, for example the SARS-CoV-2 coronavirus, unknown before it caused a COVID-19 pandemic in 2020.

Viruses evolve extremely quickly, so reliable methods for viral host prediction are necessary to safeguard biosecurity and biosafety alike. Novel human-infecting viruses are difficult to detect with standard bioinformatics workflows. Here, we predict whether a virus can infect humans directly from next-generation sequencing reads. We show that deep neural architectures significantly outperform both shallow machine learning and standard, homology-based algorithms, cutting the error rates in half and generalizing to taxonomic units distant from those presented during training. Further, we develop a suite of interpretability tools and show that it can be applied also to other models beyond the host prediction task. We propose a new approach for convolutional filter visualization to disentangle the information content of each nucleotide from its contribution to the final classification decision. Nucleotide-resolution maps of the learned associations between pathogen genomes and the infectious phenotype can be used to detect regions of interest in novel agents, for example, the SARS-CoV-2 coronavirus, unknown before it caused a COVID-19 pandemic in 2020. All methods presented here are implemented as easy-to-install packages not only enabling analysis of NGS datasets without requiring any deep learning skills, but also allowing advanced users to easily train and explain new models for genomics.

Novel pathogens evolve quickly and may emerge rapidly, causing dangerous outbreaks or even global pandemics. Next-generation sequencing is the state of the art in open-view pathogen detection, and one of the few methods available at the earliest stages of an epidemic, even when the biological threat is unknown. Analyzing the samples as the sequencer is running can greatly reduce the turnaround time, but existing tools rely on close matches to lists of known pathogens and perform poorly on novel species. Machine learning approaches can predict if single reads originate from more distant, unknown pathogens but require relatively long input sequences and processed data from a finished sequencing run. Incomplete sequences contain less information, leading to a trade-off between sequencing time and detection accuracy. Using a workflow for real-time pathogenic potential prediction, we investigate which subsequences already allow accurate inference. We train deep neural networks to classify Illumina and Nanopore reads and integrate the models with HiLive2, a real-time Illumina mapper. This approach outperforms alternatives based on machine learning and sequence alignment on simulated and real data, including SARS-CoV-2 sequencing runs. After just 50 Illumina cycles, we observe an 80-fold sensitivity increase compared to real-time mapping. The first 250 bp of Nanopore reads, corresponding to 0.5 s of sequencing time, are enough to yield predictions more accurate than mapping the finished long reads. The approach could also be used for screening synthetic sequences against biosecurity threats.

MotivationNovel pathogens evolve quickly and may emerge rapidly, causing dangerous outbreaks or even global pandemics. Next-generation sequencing is the state-of-the art in open-view pathogen detection, and one of the few methods available at the earliest stages of an epidemic, even when the biological threat is unknown. Analyzing the samples as the sequencer is running can greatly reduce the turnaround time, but existing tools rely on close matches to lists of known pathogens and perform poorly on novel species. Machine learning approaches can predict if single reads originate from more distant, unknown pathogens, but require relatively long input sequences and processed data from a finished sequencing run.ResultsWe present DeePaC-Live, a Python package for real-time pathogenic potential prediction directly from incomplete sequencing reads. We train deep neural networks to classify Illumina and Nanopore reads and integrate our models with HiLive2, a real-time Illumina mapper. DeePaC-Live outperforms alternatives based on machine learning and sequence alignment on simulated and real data, including SARS-CoV-2 sequencing runs. After just 50 Illumina cycles, we increase the true positive rate 80-fold compared to the live-mapping approach. The first 250bp of Nanopore reads, corresponding to 0.5s of sequencing time, are enough to yield predictions more accurate than mapping the finished long reads. Our approach could also be used for screening synthetic sequences against biosecurity threats.AvailabilityThe code is available at: https://gitlab.com/dacs-hpi/deepac-live and https://gitlab.com/dacs-hpi/deepac. The package can be installed with Bioconda, Docker or pip.ContactJakub.Bartoszewicz@hpi.de, Bernhard.Renard@hpi.deSupplementary informationSupplementary data are available online.

Background Emerging pathogens are a growing threat, but large data collections and approaches for predicting the risk associated with novel agents are limited to bacteria and viruses. Pathogenic fungi, which also pose a constant threat to public health, remain understudied. Relevant data remain comparatively scarce and scattered among many different sources, hindering the development of sequencing-based detection workflows for novel fungal pathogens. No prediction method working for agents across all three groups is available, even though the cause of an infection is often difficult to identify from symptoms alone. Results We present a curated collection of fungal host range data, comprising records on human, animal and plant pathogens, as well as other plant-associated fungi, linked to publicly available genomes. We show that it can be used to predict the pathogenic potential of novel fungal species directly from DNA sequences with either sequence homology or deep learning. We develop learned, numerical representations of the collected genomes and visualize the landscape of fungal pathogenicity. Finally, we train multi-class models predicting if next-generation sequencing reads originate from novel fungal, bacterial or viral threats. Conclusions The neural networks trained using our data collection enable accurate detection of novel fungal pathogens. A curated set of over 1400 genomes with host and pathogenicity metadata supports training of machine-learning models and sequence comparison, not limited to the pathogen detection task. Availability and implementation The data, models and code are hosted at https://zenodo.org/record/5846345, https://zenodo.org/record/5711877 and https://gitlab.com/dacs-hpi/deepac. Supplementary information Supplementary data are available at Bioinformatics online.

Motivation:We expect novel pathogens to arise due to their fast-paced evolution, and new species to be discovered thanks to advances in DNA sequencing and metagenomics. Moreover, recent developments in synthetic biology raise concerns that some strains of bacteria could be modified for malicious purposes.Traditional approaches to open-view pathogen detection depend on databases of known organisms, which limits their performance on unknown, unrecognized, and unmapped sequences. In contrast, machine learning methods can infer pathogenic phenotypes from single NGS reads, even though the biological context is unavailable. Results: We present DeePaC, a Deep Learning Approach to Pathogenicity Classification. It includes a flexible framework allowing easy evaluation of neural architectures with reverse-complement parameter sharing. We show that convolutional neural networks and LSTMs outperform the state-of-the-art based on both sequence homology and machine learning. Combining a deep learning approach with integrating the predictions for both mates in a read pair results in cutting the error rate almost in half in comparison to the previous state-of-the-art. Availability: The code and the models are available at: https://gitlab.com

Summary The ongoing pandemic caused by SARS-CoV-2 emphasizes the importance of genomic surveillance to understand the evolution of the virus, to monitor the viral population, and plan epidemiological responses. Detailed analysis, easy visualization, and intuitive filtering of the latest viral sequences are powerful for this purpose. We present CovRadar, a tool for genomic surveillance of the SARS-CoV-2 Spike protein. CovRadar consists of an analytical pipeline and a web application that enable the analysis and visualization of hundreds of thousands sequences. First, CovRadar extracts the regions of interest using local alignment, then builds a multiple sequence alignment, infers variants and consensus, and finally presents the results in an interactive app, making accessing and reporting simple, flexible and fast. Availability and implementation CovRadar is freely accessible at https://covradar.net, its open-source code is available at https://gitlab.com/dacs-hpi/covradar. Supplementary information Supplementary data are available at Bioinformatics online.