January 12, 2023  |  Infectious disease research

Meeting the threat of antimicrobial resistance with HiFi sequencing and artificial intelligence

 

After conducting a battery of technology comparisons, OpGen’s Ares Genetics has adopted PacBio HiFi microbial whole genome sequencing. The company believes the read length and high accuracy of HiFi data will enable it to develop a suite of cutting-edge AI solutions with potential to revolutionize the way we combat the spread of antimicrobial resistance.
 

An urgent risk to global health


Before the discovery and mass production of antimicrobial compounds, seemingly innocuous microbial infections were a common cause of human mortality. Today, public health experts around the world are concerned that this could once again become the norm due to the proliferation of antimicrobial resistance (AMR).

Thankfully, some of the top minds in microbiology, medicine, and public health are tackling this problem head-on with important changes being made in everything from patient care processes to livestock production. As researchers continue to uncover the complexities of antimicrobial resistance, it has become clear that there is a critical need for public-health solutions that integrate molecular biology, microbiology, and artificial intelligence (AI). This is exactly the approach being taken by the team at Ares Genetics, an OpGen group company, that aims to revolutionize how we address the global challenge of AMR. OpGen’s ARES portfolio includes services and solutions for outbreak surveillance, infection control, AMR prediction, and the targeted delivery of antimicrobial therapy.
 

A state-of-the-art solution fueled by PacBio long-read sequencing


OpGen’s ARES suite of sophisticated machine learning algorithms can comprehensively determine the genetic profiles of pathogens and use that information to not only detect markers of AMR but predict how an infection will respond to an antibiotic. The potential for ARES technology to massively increase the speed and efficiency with which public health authorities can detect, monitor, and combat AMR is tremendous. To embrace this potential, OpGen with its ARES suite of offerings, has taken a laser-like focus on the quality of the data being used to train their powerful AI systems.

Most next generation sequencing (NGS) technologies currently available to guide such systems are reliant on extensive computational processing to reconstruct genes and genomes by stitching together large numbers of very short DNA sequences. This complex data processing requirement means that information obtained using conventional short-read sequencing may often contain mis-assemblies and truncations which can interfere with gene detection and lead to potential false-negatives when looking for the presence of AMR genes in pathogens. Moreover, these conventional NGS approaches are unable to reliably determine whether AMR-associated genes are located on the native chromosome of the subject microbe or on a plasmid, a small piece of circularized DNA that can be freely exchanged between cells and across species. Additionally, some long-read technologies can miss important AMR genes and their functions due to noisy data outputs and underlying inaccuracy. The ability to effectively capture plasmid sequences and correctly determine if they contain AMR genes is critical, as their presence indicates that resistance could spread quickly within a hospital or other locality. To avoid the potential pitfalls of short reads and less accurate long-read sequencing technologies for AMR outbreak monitoring applications, OpGen’s ARES team has begun using PacBio HiFi microbial whole genome sequencing to give their proprietary AI the highest quality data possible.

With lengths of more than 10,000 bases per read and impeccable accuracy, HiFi sequencing excels where competing technologies fall short. HiFi whole genome assemblies do not require the complex software workarounds that define short-read NGS and similarly, HiFi data does not require the extensive data filtering that plagues other noisy, error-prone long-read solutions. For this reason, HiFi data can be used to assemble bacterial genomes that exceed the quality of reference standards and can deftly capture plasmids.

To put these strengths into practice and to verify that the benefits of HiFi sequencing translate to a measurable improvement in AMR detection, OpGen’s ARES team compared PacBio HiFi sequencing to Oxford Nanopore Technologies (ONT) and Illumina using type-strains of seven AMR bacteria. In a test for AMR marker detection, PacBio HiFi data matched or exceeded ONT and Illumina in accuracy, sensitivity, and precision for all seven bacterial species tested.
 

ACCURACY SENSITIVITY PRECISION
Species  Sample size (n) Illumina ONT SMRT Link Hifiasm Illumina ONT SMRT Link Hifiasm Illumina ONT SMRT Link Hifiasm
A. baumanii 186 0.989 0.978 1 1 0.995 0.978 1 1 0.995 1 1 1
P. aeurignosa 334 0.991 0.985 0.997 0.997 0.994 0.988 1 1 0.997 0.997 0.997 0.997
K. pneumoniae 183 0.989 0.973 1 1 0.989 0.978 1 1 1 0.994 1 1
S. aureus 117 0.983 0.983 1 1 0.983 0.983 1 1 1 1 1 1
E. coli 286 0.983 0.976 0.986 0.99 0.982 0.979 0.989 0.993 1 0.986 0.996 0.996
E. faecium 29 0.759 0.793 0.828 0.862 0.957 0.957 1 1 0.786 0.815 0.821 0.852
E. colacae 98 0.99 0.959 0.99 0.969 0.989 0.989 0.989 0.979 1 0.989 1 0.989

 
Ares scientists went on to compare these three sequencing technologies in a culture-based antimicrobial sensitivity test wherein ARES AI algorithms were fed the whole genome assemblies created with each technology and were then asked to predict a bacterium’s resistance or susceptibility to a specific antimicrobial compound. The organisms were then cultured and challenged with each antimicrobial compound using standardized broth microdilution antimicrobial sensitivity testing (AST) and the results were compared.
 
 
RESULT
Illumina SBS
(SPAdes v3.15.2)
ONT R10.4
(Flye v2.9 + Medakka v 1.6.1)
PacBio HiFi
(SMRT Link)
PacBio HiFi
(Hifasm)
Categorical Agreement 90.90% 93.20% 95.50% 95.50%
Balanced Accuracy  88.30% 90.00% 93.30% 93.30%
Major Error 3.40% 0% 0% 0%
Very Major Error 20.00% 20.00% 13.30% 13.30%

(Major Error is the percentage of false-resistant results over all phenotypic susceptible results. Very Major Error is the percentage of false-susceptible results over all phenotypic resistant results).
 
Here again, PacBio HiFi assemblies outperformed Illumina and ONT with results that agreed with the AST results the most. Though preliminary, these results demonstrate that for disease monitoring and public health applications, where every detail matters, HiFi sequencing is clearly the technology of choice.
 

Your partner in the fight against antimicrobial resistance


At PacBio, we are committed to being a reliable partner who supports and inspires your mission to safeguard and improve human health. If you have further questions about how PacBio sequencing can be used to combat antimicrobial resistance or other pressing public health challenges, connect with a PacBio scientist today.
 

Interested in reading more?

Public health + surveillance

A new standard in pathogen genomics

Infectious disease research

Talk with an expert

If you have a question, need to check the status of an order, or are interested in purchasing an instrument, we're here to help.