Journal of Biomedical Engineering and Biosciences (JBEB)

ISSN: 2564-4998

1. Introduction

The global rise of antibiotic resistance is a growing threat to food safety and environmental health. With treated municipal wastewater increasingly reused for agricultural irrigation, resistant bacteria and antibiotic resistance genes (ARGs) have been shown to move through interconnected environmental compartments like soil, plants, and water systems, posing serious and complex challenges for tracking and mitigation [1–4]. Recent studies show that some nonantibiotic pesticides can lead to the development of antibiotic-resistant bacteria and cross-resistance to multiple antibiotics, highlighting the complex nature of resistance development in agricultural environments [5–8]. Foundational studies in environmental microbiology established wastewater as a critical reservoir of ARGs, where the vast gene exchange among microbial populations was revealed through metagenomic and genomic investigations. [9-12]. The lack of new antibiotics, alongside the rise of pan-resistant and multiresistant pathogens, underscores the urgency of improved surveillance and management in environmental contexts [13–15]. ARGs are widespread across multiple environmental components, and understanding their persistence and dissemination is crucial for effective environmental tracking and risk mitigation [16–18].

As sequencing became more accessible, genome analysis pipelines were developed to identify ARGs and characterize bacterial samples with greater accuracy and precision. Tools such as TORMES, Nullarbor, and nf-core/mag have contributed as frameworks for metagenomic assembly. However, these platforms often present challenges for large-scale or non-clinical datasets due to limited modularity. These frameworks also have steep learning curves and high computational demands [19–21]. Bactopia, a more flexible and customizable pipeline, integrates numerous tools for quality control, assembly, annotation, and resistance gene detection. This makes Bactopia a particularly suitable pipeline for bacterial genome analysis across different environments [22 - 24]. Despite its advantages, manual execution of Bactopia across multiple datasets remains time-consuming, error-prone, and inefficient when monitoring ARGs in complex agricultural systems.

Numerous studies have characterized ARGs in wastewater systems, but few implement scalable, automated computational workflow methods that are designed for ARG tracking across various environments in addition to comparisons among genome analysis pipelines. Additionally, the differences in genome assemblers and their impact on ARG detection accuracy have not been comprehensively evaluated.  These limitations hinder the development of efficient approaches to track antibiotic resistance surveillance.

To address these limitations, a Python-based automation framework was developed to streamline Bactopia execution. Bactopia execution, using Docker integration, was able to deal with errors and automate data formatting. This system reduces manual efforts and allows entire directories of FASTQ files to be processed through a single command, increasing reproducibility and scalability [25,26]. The workflow was applied to Escherichia coli strains isolated from the secondary effluent of a local municipal wastewater treatment plant.

In addition to automation, comparative evaluations were performed between Bactopia, Nullarbor, TORMES, and nf-core/mag to assess performance, modularity, and accuracy in detecting ARGs in environmental samples. The study further analyzed the performance of assemblers within Bactopia (SPAdes, SKESA, and MEGAHIT) to determine optimal assembly strategies for ARG detection in paired-end reads. Finally, visualization was implemented in Python using Pandas, Matplotlib, Seaborn, and Holoviews to generate heat maps, multiple sequence alignments, and chord diagrams. Visualization allowed for easier and convenient interpretation of ARG patterns and resistance pathways across environmental systems [29–31]. ARGs can persist and disseminate across environmental matrices, and strategies for their detection and tracking are essential for mitigating public health risks [32,33].

By integrating automation, comparative benchmarking, and visualization, this work establishes a genomic and computational framework for tracking antibiotic resistance gene spread in wastewater systems using treated effluent. The approach addresses key methodological gaps in the current work done to address ARG transmission and monitoring while supporting reproducibility, accessibility, and scalability. It also supports better environmental surveillance in the future, as well as assessing risks and aiding mitigation efforts.

2. Genome Data

Genomes of antibiotic-resistant E. coli were used to test the analysis pipeline. Streptomycin-resistant E. coli strains were isolated from the secondary effluent of a local municipal wastewater treatment plant. Selective plates were prepared using ECD ChromoSelect Agar with MUG (Sigma-Aldrich) containing 32 mg/L streptomycin, which corresponds to the resistance breakpoint defined by the National Antimicrobial Resistance Monitoring System. Resistant E. coli appear as blue colonies, which were picked, revived in LB medium, and archived in glycerol for downstream analysis. To evaluate pipeline performance across a range of resistance levels, we selected isolates exhibiting streptomycin resistance fold changes of 2, 4, 8, 16, 32, and 64 for whole-genome sequencing. Genomic DNA was extracted using the Qiagen PowerSoil Pro Kit, and sequencing was performed on an Illumina MiSeq platform (2 × 150 bp) at SeqCenter, LLC.|

3. Systems Setup

3. 1. Automation System Development

A Python-based code was developed to construct and run Docker-based commands, automating Bactopia execution. The script required a single input and a URL to a directory containing raw FASTQ files, making it efficient for large-scale processing. Before execution, Shutil verified Bactopia’s availability, followed by a subprocess executing the command, with error handling to identify any failures. After execution, code utilizing Python’s pandas module formatted the data for readability. The pipeline was applied to bacteria isolated from secondary effluent in a local municipal wastewater treatment plant. It could also be applied to investigate samples from a more complex water-soil-plant continuum in water reuse applications to provide a comprehensive view of ARG transmission across environmental compartments.

3.2 Assembler Comparison Setup

Bactopia assemblers were then evaluated and compared for performance, accuracy, and scalability to identify the most effective tools for detecting ARGs in genomic datasets. Treated municipal wastewater was sampled to trace resistance dissemination. The whole bacterial genome was sequenced to identify key genes conferring the resistance, and a program was developed to automate Bactopia runs for each of the eight genomic datasets. The data was processed using SPAdes, Skesa, and Megahit assemblers and formatted to sample certain attributes. SPAdes was used for its high accuracy in assembling small bacterial genomes, while SKESA was selected for its speed and conservative assembly approach. In order to handle potential contaminants, MEGAHIT, originally designed for metagenomic datasets, was included. SPAdes, Bactopia’s default assembler, was used as a reference to compare the runtime of the three assemblers

3. 3. Visualization System Development

The program was implemented in Python due to its support for data visualization and processing libraries, making the language well-suited for handling biological datasets. Code was constructed to support three forms of visual output. Data is inputted in CSV format and parsed using Pandas.  Then, separate modules generate visualizations using Matplotlib, HoloViews, and Seaborn. These libraries produced three types of graphic visualization in the form of heat maps, multiple sequence alignment (MSA), and interactive chord graphs.

3. 4. Pipeline Comparison

Several genome analysis pipelines were evaluated for their performance in processing environmental samples that model wastewater irrigation conditions. To obtain relevant microbial communities, we collected and isolated antibiotic-resistant bacteria from wastewater systems. Then, we analyzed the samples using four bioinformatics pipelines: Bactopia, Nullarbor, TORMES, and nf-core/mag. Each pipeline was tested on the same set of samples to compare their capabilities in assembly, annotation, and detection of resistance elements.

4. Results and Discussions

4.1 Automation of the Bactopia Pipeline

The automated Bactopia pipeline successfully detected antibiotic resistance genes (ARGs) in E. coli isolates. Automation reduced manual input time drastically, enabling rapid processing of large environmental samples. A wide range of ARGs was identified, and formatted data allowed straightforward and convenient comparison across all strains. Heat maps, multiple sequence alignments, and chord graphs highlighted ARG movement and distribution patterns.

The Python-based code streamlined execution by constructing and running Docker commands, verifying Bactopia availability, and formatting outputs using Pandas. It also handled potential errors during genome analysis. The data was processed using SPAdes, SKESA, and MEGAHIT assemblers, part of Bactopia, and formatted to highlight specific attributes. SPAdes produced high-quality assemblies, SKESA offered faster runtimes with slightly more fragmented assemblies, and MEGAHIT effectively handled potential contaminants while requiring higher computational resources. For general purposes, SPAdes offers the best balance of efficiency and quality.

Compared with manual analysis, the automated pipeline eliminated repeated command execution and file management. It reduced the possibility of human error and ensured consistent results run after run. By skipping intermediate steps, the workflow allows direct processing from raw sequencing data to ARG classification and detection. This greatly improves efficiency in analyzing genomic data for ARGs. Figure 1 illustrates this workflow comparison, showing the traditional manual approach versus the streamlined automated pipeline.

The automated Bactopia pipeline improved speed for ARG detection. It provides a scalable framework for monitoring antibiotic resistance in agricultural systems using treated wastewater and can be adapted for broader environmental surveillance and risk assessment.

4.2 Bactopia’s Performance in Comparison to Other Comparable Assemblers

The performance of SPAdes, SKESA, and MEGAHIT was assessed to understand how each assembler supports the detection of antibiotic resistance genes (ARGs) in environmental genomic datasets, emphasizing how well each tool captured biologically significant ARG information in complex wastewater-derived E. coli samples. Multiple metrics were considered, including runtime efficiency, assembly quality, and ability to handle complex datasets of streptomycin-resistant E. coli from wastewater effluent. SPAdes consistently produced more coherent and biologically interpretable assemblies, using fewer contigs and providing more complete representations of resistance genes. This structural continuity made ARG profiles easier to reconstruct and interpret, which becomes crucial when analyzing environmental samples with heterogeneous genetic backgrounds. This makes it well-suited for reconstructing complete ARG profiles. SKESA offered speed and computational efficiency at the cost of increased amounts of fragmentation. The resulting assemblies often broke ARG regions into smaller pieces, limiting resolution and obscuring gene context when identifying resistance mechanisms in complex samples. MEGAHIT was effective at handling large datasets with potential contaminants, showing strength and robustness in metagenomic settings. However, its practicality is hindered by its high computational demands and longer runtimes, limiting its use for targeted ARG analysis. Overall, while each assembler offered distinct advantages for targeted use, SPAdes provided the most reliable balance between assembly coherence and biological insight, making it the most effective choice.

Figure 2 illustrates the relative runtime efficiency of each assembler. SPAdes was set as the baseline at 129 minutes (100% efficiency), with SKESA requiring 141 minutes (91% efficiency) and MEGAHIT using 135 minutes (88% efficiency). This visual comparison highlights the trade-offs between speed and assembly quality as SPAdes delivers higher-quality assemblies with slightly longer runtime, SKESA prioritizes speed with minor fragmentation, and MEGAHIT provides robust assembly for complex metagenomic data at the cost of computational demand. Standardized formatting of output data for desired traits enabled direct comparison of ARG profiles across all samples, allowing for clearer identification of gene classes, resistance clusters, and potential environmental dissemination patterns.

These results emphasize the importance of assembler selection when monitoring antibiotic resistance in water reuse systems. Choosing the right assembler can directly impact the accuracy and completeness of ARG detection, especially in complex environmental samples. SPAdes emerged as the most reliable option for genome assembly, while SKESA and MEGAHIT offer alternatives when computational speed or metagenomic diversity is prioritized. Optimizing assembler choice ensures scalable and reproducible analysis, which supports more effective monitoring and assessment of ARG dissemination.

4.3 Development of an Automated Data Visualization

The automated Bactopia-based Python pipeline enabled high-throughput analysis and visualization of ARGs, which transformed genomic data into interpretable visual outputs. Ultimately,  these visualization techniques, including MSAs, a heatmap, and a chord map, facilitate the exploration of resistance gene dissemination across the samples tested.

1. MSAs (Figures 3-4): Figure 3 specifically focuses on the arcF gene across eight samples, with non-consensus nucleotides highlighted. These MSAs were developed using the Pandas and Seaborn libraries. This type of representation provides an in-depth view of mutations within individual genes, which ultimately contribute to resistance heterogeneity, a serious public health concern. The presence of non-consensus nucleotides reflects point mutations and sequence variation, indicating that while arcF is conserved, specific positions may be under selective pressure. In Figure 4, a full-sample MSA is shown across all the E. coli samples. This type of visualization illustrates the broader genomic landscape across all the samples, which enables conclusions such as core resistance determinants, genomic diversity, and overall comparison of the samples depending on their respective sources. Conserved regions observed across all samples reflect shared resistance genes, whereas variable regions point to sample-specific adaptation or horizontal gene transfer events. This alignment technique is also beneficial to understanding how ARGs develop as a result of selective pressures, ultimately making it easier to track ARG dissemination across different environments.
2. Heat Map (Figure 6): This heat map summarizes the presence and distribution of ARGs across all eight samples. This heat map was programmed with Pandas, Seaborn, and Matplotlib libraries. Each cell represents whether or not a particular ARG is found within the sample, with dark blue indicating prevalence and yellow indicating zero prevalence. This type of visualization is crucial to the identification of resistance determinants across all the samples analyzed. Patterns such as critical resistance hotspots and the most frequently present ARGs, which pose a greater threat, can be determined. ARGs consistently detected across multiple samples likely represent resistance determinants, while ARGS restricted to individual samples indicated localized acquisition or specific selective pressures.

3. Chord map (Figure 5): This map illustrates the interconnected nature of ARGs across all eight samples. It was programmed using the Pandas and HoloViews libraries to create an interactive platform for more user efficiency. Each color corresponds to a respective sample, making it easier to see the overlaps between them. Each line reveals patterns such as co-occurrence of ARGs, gene clusters that often appear together, or shared selective pressures. Frequent co-occurrence of specific ARGs across multiple samples suggests linkage within gene clusters or elements, demonstrating potential mechanisms for coordinated resistance dissemination. This visualization technique reveals the presence of ARGs throughout various microbial communities, including their shared genes, and highlights interconnected agricultural environments that are prone to the spread of antibiotic resistance.

Together, these visualizations are essential parts of understanding patterns derived from complex genomic data. Some examples of potential patterns that could be revealed include ARG diversity, co-occurrence, and mutation-driven adaptation. The visual outputs demonstrate how sequence variation, ARG prevalence, and gene co-occurrence collectively contribute to resistance heterogeneity across the sampled environment. This type of approach is vital to the formation of these conclusions, in addition to providing a more scalable, big-picture framework for monitoring antibiotic resistance throughout different environments.

4.4 Comparing Bactopia to Other Pipelines

We tested three other pipelines and evaluated their benefits and flaws as follows:

1. Nullarbor (Figure 7): Nullarbor detected the presence of ARGs within seven genomes. It was simple to install and had a user-friendly interface. It produced consistent outputs but required long runtimes and had limited flexibility for non-clinical datasets.
2. TORMES (Figure 8): TORMES detected ARGs (across all eight genomes), the contig where the gene was found, and where the gene was located within the contig. It was lightweight and relatively simple to configure with paired-end FASTQ files; however, it lacked the scalability and modularity needed for complex environmental samples.
3. Nf-core/mag (Figure 9): Nf-core/mag is commonly used by the NIH, and was executed through Nextflow with a custom configuration. For eight genomes, it delivered lengths of assembled sequences (contigs), displaying robust metagenomic assembly capabilities but with the steepest learning curve and the highest computational demands. Several reads were not long enough, so the pipeline skipped over them, resulting in only four contigs.

The Bactopia pipeline delivered the best results due to its highly customizable and modular framework, which is ideal for bacterial isolate analysis. The other pipelines are designed for isolated or well-characterized samples, limiting their effectiveness in our study.

Bactopia outperformed Nullarbor, TORMES, and Nf-core/mag in processing our datasets. It is particularly useful for versatile, complex genome analysis, so it performed the best out of all the other pipelines tested. The work compared leading genome analysis pipelines, improved ARG monitoring, and can be adapted for resistance monitoring, risk assessment, and mitigation.

5. Conclusions

A Python program was developed to automate the Bactopia pipeline, making the system streamlined and efficient. The automated pipeline significantly decreased the time needed to initialize analysis and standardized data formatting. Several visualization programs were additionally developed, providing a more readable format of data.

Assemblers, including MEGAHIT, SPAdes, and SKESA, were compared based on runtime efficiency and quality of assembly. SPAdes was the most efficient assembler with the shortest runtime and consistently produced high-quality genome assemblies with fewer contigs and higher N50 values. This suggests that SPAdes may be more reliable for reconstructing complete ARG profiles, particularly in complex environmental samples.  

Three pipelines other than Bactopia were tested on the same set of data: Nullarbor, Tormes, and nf-core/mag. Bactopia delivered a highly customizable and modular framework ideal for bacterial isolate analysis.  Nullarbor offered a user-friendly interface but required longer runtimes and was less flexible with non-clinical datasets. TORMES was lightweight and relatively simple to configure with paired-end FASTQ files; however, it lacked the scalability and modularity needed for complex environmental samples. Nf-core/mag was executed through Nextflow with a custom configuration. It delivered robust metagenomic assembly capabilities with the steepest learning curve and the highest computational demands. Overall, Bactopia outperformed TORMES, Nullarbor, and nf-core/mag in processing the dataset.

The automation and visualization programs provided an efficient, easy-to-use system for analyzing genomic data using the Bactopia pipeline. The framework can be adapted for broader surveillance of antibiotic resistance in diverse environmental contexts, reducing manual input time and standardizing output formatting. Its modular structure also supports increased scalability, enabling the processing of larger sample sets or integrating additional datasets without major restructuring. Our comparison of assemblers and pipelines offers insight into selecting important parameters for analysis. Future work may explore hybrid approaches or assembler-specific optimization to improve the detection of resistance genes across diverse sample types.

Acknowledgements

We would like to acknowledge staff from the Western Riverside County Regional Wastewater Authority for providing the activated sludge samples. This study was supported by the National Science Foundation (Award No. 2045658, for C.Z. and Y.M.).

We thank a gift from Beckman Laser Institute Inc. to LS. Special thanks to the private donors to our UCSD IEM BTC center: Dr. Shu Chien from UCSD Bioengineering, Dr. Lizhu Chen from CorDx Inc., Dr. Xinhua Zheng, David & Leslie Lee, Mingwei Hu and Wei Shi for their generous donations.

References