Biology subjects

Andreopoulos, W. B.

Publications and source records attributed to Andreopoulos, W. B..

5 recordsLinked to original sources

bioRxiv2026

Characterization of artificial riboswitches for Coxsackievirus B3 detection using machine learning

With the increase of recycled water to offset water demand, the potential possibility to spread contagious RNA viruses, such as Coxsackievirus B3, increases. However, detection of viral particles remains challenging because of low viral concentrations in wastewater and high mutation rates of the RNA virus. Robust monitoring is needed with low cost and low infrastructure technologies to increase accessibility of monitoring technologies worldwide. Novel viral detection methods have been developed, based on synthetic riboswitches that bind to the target virus and trigger a reporter gene, thus amplifying the detection signals. Such monitoring technologies can leverage machine learning to optimize the candidate nucleic acid sequences used for detection. To support the design of effective riboswitches, we present a machine learning model for classifying riboswitch performance, integrating RNA sequence data with secondary structural features based on free energy calculations and parameters that represent single strandedness. This model uses a sparsely gated Mixture of Experts (MoE) architecture to route sequence and thermodynamic features to specialized experts, achieving strong generalization performance across cross-validation and held-out testing. When evaluated against baseline Decision Tree, Random Forest, Gaussian Naive Bayes, and Dense Neural Network classifiers, the MoE model demonstrated near-zero classification errors. Furthermore, post-hoc feature importance and k-mer analyzes reveal that near-perfect predictive performance in silico is strongly driven by sequence constructs in addition to generalizable folding mechanics. An ablation study of numerical feature and sequence inputs showed that while the Dense Neural Network also achieved high accuracy across all ablation conditions, the MoE architecture was retained as the primary model because its specialized subnetworks do not require all model parameters to be activated offering a potential computational advantage.

bioinformatics↗

bioRxiv2026

Autonomous Spatial Transcriptomics Analysis (ASTA): Demonstrating Performance Improvements through Clustering, Biological Annotation, and AI-Driven Discovery

Spatial transcriptomics keeps measurement of gene expression while preserving spatial context, yet traditional analysis methods face challenges in computational efficiency, biological interpretability, and autonomous discovery. This project presents a framework solving these issues through three parts: (1) an ensemble clustering system achieving 66.7% improvement over baseline average and 23.9% over best single method with silhouette score of 0.540 and statistical significance (p = 0.0032, Cohens d = 1.82); (2) a knowledge-based clustering framework that annotates 88.6% of cells across 8 ovarian cell types using 428 marker genes; and (3) a GPT-4o-mini-powered autonomous agent that generated 3 biological hypotheses with validations.

bioinformatics↗

bioRxiv2022

On the synergies between ribosomal assembly and machine learning tools for microbial identification

Genome assembly tools are used to reconstruct genomic sequences from raw sequencing data, which are then used for identifying the organisms present in a metagenomic sample. More recently, machine learning approaches have been applied to a variety of bioinformatics problems, and in this paper, we explore their use for organism identification. We start out by evaluating several commonly used metagenomic assembly tools, including PhyloFlash, MEGAHIT, MetaSPAdes, Kraken2, Mothur, UniCycler, and PathRacer, and compare them against state-of-the art deep learning-based machine learning classification approaches represented by DNABERT and DeLUCS, in the context of two synthetic mock community datasets. Our analysis focuses on determining whether ensembling metagenome assembly tools with machine learning tools has the potential to improve identification performance relative to using the tools individually. We find that this is indeed the case, and analyze the level of effectiveness of potential tool ensembling for organisms with different characteristics (based on factors such as repetitiveness, genome size, and GC content). Author SummaryMetagenomic studies focus on the challenging problem of identifying the presence and abundance of different species in a sample. This process typically involves the creation of digital reads from the sample which correspond to small parts of the genome sequence, and then have to be assembled together by a genome assembly tool. More recently, machine learning approaches have been applied to a variety of bioinformatics problems, and in this paper, we explore their use for organism identification, and how they might complement traditional bioinformatics approaches. We conduct experiments with two representative state-of-the-art machine learning approaches and six metagenomic assembly tools in the context of two synthetic datasets. We find that for organisms with certain characteristics (levels of repetitiveness, GC content, and genome size), ensembling metagenome assembly tools with machine learning tools has the potential to improve species identification performance relative to using the tools individually.

bioinformatics↗

bioRxiv2021

Deeplasmid: Deep learning accurately separates plasmids from bacterial chromosomes

Plasmids are mobile genetic elements that play a key role in microbial ecology and evolution by mediating horizontal transfer of important genes, such as antimicrobial resistance genes. Many microbial genomes have been sequenced by short read sequencers and have resulted in a mix of contigs that derive from plasmids or chromosomes. New tools that accurately identify plasmids are needed to elucidate new plasmid-borne genes of high biological importance. We have developed Deeplasmid, a deep learning tool for distinguishing plasmids from bacterial chromosomes based on the DNA sequence and its encoded biological data. It requires as input only assembled sequences generated by any sequencing platform and assembly algorithm and its runtime scales linearly with the number of assembled sequences. Deeplasmid achieves an AUC-ROC of over 93%, and it was much more precise than the state-of-the-art methods. Finally, as a proof of concept, we used Deeplasmid to predict new plasmids in the fish pathogen Yersinia ruckeri ATCC 29473 that has no annotated plasmids. Deeplasmid predicted with high reliability that a long assembled contig is part of a plasmid. Using long read sequencing we indeed validated the existence of a 102 Kbp long plasmid, demonstrating Deeplasmids ability to detect novel plasmids. AvailabilityThe software is available with a BSD license: deeplasmid.sourceforge.io. A Docker container is available on DockerHub under: billandreo/deeplasmid. Contactwandreopoulos@lbl.gov alevy@mail.huji.ac.il

bioinformatics↗

bioRxiv2020

The extracellular contractile injection system is enriched in environmental microbes and associates with numerous toxins

Bacteria employ toxin delivery systems to exclude bacterial competitors and to infect host cells. Characterization of these systems and the toxins they secrete is important for understanding microbial interactions and virulence in different ecosystems. The extracellular Contractile Injection System (eCIS) is a toxin delivery particle that evolved from a bacteriophage tail. Four known eCIS systems have been shown to mediate interactions between bacteria and their invertebrate hosts, but the broad ecological function of these systems remains unknown. Here, we identify eCIS loci in 1,249 prokaryotic genomes and reveal a striking enrichment of these loci in environmental microbes and absence from mammalian pathogens. We uncovered 13 toxin genes that associate with eCIS from diverse microbes and show that they can inhibit growth of bacteria, yeast or both. We also found immunity genes that protect bacteria from self-intoxication, supporting an antibacterial role for eCIS. Furthermore, we identified multiple new eCIS core genes including a conserved eCIS transcriptional regulator. Finally, we present our data through eCIStem; an extensive eCIS repository. Our findings define eCIS as a widespread environmental prokaryotic toxin delivery system that likely mediates antagonistic interactions with eukaryotes and prokaryotes. Future understanding of eCIS functions can be leveraged for the development of new biological control systems, antimicrobials, and cell-free protein delivery tools.

microbiology↗