Original papers
Multidrug resistance analysis method for pathogens of cow mastitis based on weighted-association rule mining and similarity comparison

https://doi.org/10.1016/j.compag.2021.106411Get rights and content

Abstract

Mastitis is one of the most common diseases and causes the greatest economic loss in dairy farming. Antibiotics are the most effective drugs to prevent and treat bacterial infection of mastitis. However, yet the growing problem of drug resistance, especially multidrug resistance (MDR), poses a great threat to disease control. To understand the MDR rules in bacteria of cow mastitis from national level, the main bacteria from cows with mastitis in large-scale farms were isolated and identified in China, and then drug sensitivity tests were conducted to establish a drug resistance data set. Aiming at the problem of numerous and disordered drug resistance data and lack of extensive correlations, a weighted Apriori association rule mining algorithm in conjunction with the bacterial drug resistance prevalence is proposed. We analyzed the associations between different antibiotics of key bacteria, extracted and visualized the key trends of high resistance prevalence and frequent occurrence, and discovered MDR patterns. Finally, a similarity comparison method based on Euclidean measurement was proposed to compare the relative MDR rules of different bacteria from the overall level with support, confidence, and promotion as characteristic parameters. The drug resistance data set showed that staphylococcus were the main bacteria isolated from dairy cow mastitis in China. Then based on the association rule algorithm, the important rules between different antibiotics resistance in this dataset were identified. In addition, the MDR patterns of different bacteria were visualized and analyzed by using the chord diagram. The results showed the bacteria are highly resistant to penicillin, gentamicin, and ampicillin, and most other antibiotics were linked with these three antibiotics. Finally, the high correlations and main rules in different bacteria were confirmed by a similarity comparison method. The assessment model and conclusions of this study are potentially valuable for assessing the evolution of MDR patterns, providing a scientific basis for relevant authorities to guide the rational use of antibiotics in the farming industry.

Introduction

The high rate and irregular use of antibiotics make the drug resistance of bacteria increasingly serious, which has become one of the most serious problems in the world (Clifford et al., 2018, Davies and Wales, 2019), especially the problem of multidrug resistance (MDR) (Van Boeckel et al., 2019). At present, animal antibiotic consumption accounts for more than half of the total consumption of antibiotics, significantly affecting the prevention and treatment of animal diseases (Van Hecke et al., 2017), and the drug-resistant bacteria produced by farming can spread to the population through the food chain or to the environment through excrement, posing a great threat to human health (Hudson et al., 2017, Zhang et al., 2015). The government has taken strong measures to solve the drug resistance problem (Hvistendahl, 2012, Li et al., 2020), but currently mainly focuses on hospital and meat products such as poultry and pigs, with insufficient attention to dairy products (Xu et al., 2020).

Cow mastitis is one of the most frequent and damaging diseases in dairy cows (Motaung et al., 2017, Rajamanickam et al., 2019), in which pathogenic bacterial infection is the most leading cause. In MDR, pathogenic bacteria are resistant to two or more antibiotics, which is a great threat and seriously affects the effective prevention and treatment of cow mastitis (Cheng and Han, 2020). Understanding the patterns of MDR in specific bacteria can help to guide antibiotic drug management and therapy, and it is also important for the rational use of antibiotics and prevention and control of animal diseases on farms (Ludwig et al., 2013).

The current research on drug resistance and MDR is still inadequate (Sommer et al., 2017). Studies mainly focused on the bioinformatics of genes and structural proteins (Su et al., 2019, Thomas et al., 2015, Zhong et al., 2011) and the mechanism of action of antibiotics (Gao et al., 2019, Yang et al., 2019). Although it is possible to obtain MDR resistance genes and propose new solutions to solve the problem of drug resistance from a microscopic and pharmacology perspective, this approach is time-consuming to implement, and it is difficult to evaluate the overall use of antibiotics to guide dosing on farms.

The main methods for MDR analysis from the macrolevel by data analysis include Markov network, Bayesian network, and association rule mining algorithms (Safdari et al., 2020). The Bayesian network was used to estimate the interaction and mine the MDR pattern of E. coli and have identified three-way and four-way interactions between resistances (Ludwig et al., 2013), but it is an oriented graph, which only reflects the probability of bacterial emergence of resistance. Therefore, it is difficult to analyze the resistance association among multiple antibiotics. Furthermore, although the Markov network can analyze the drug resistance data well, it is statistically independent in each matter and not suitable for dealing with large amounts of data.

Association rule mining algorithms can effectively analyze sparse bacterial drug resistance data to obtain the strength of the relationship between different MDR (Agrawal and Srikant, 1994). At present, it has been used to mine significant relationships in public health and medical datasets (Lamma et al., 2003, Ma et al., 2003). Besides, in terms of drug resistance in animal-derived pathogenic bacteria, the MDR patterns of E. coli to 15 antibiotics in chickens from 2004 to 2012 based on association rule mining algorithm were analyzed, which is the first analysis of MDR patterns in the E. coli of chicken origin by using association rule mining algorithms (Cazer et al., 2019).

However, the prevalence of antibiotic resistance is one of the most important standards for evaluating the risk of resistance (Collignon et al., 2018), while traditional association rule algorithms pay more attention to the frequency of antibiotics, ignoring the differences between the prevalence of drug resistance. Therefore, some invalid rules with high support and low prevalence rate have been more easily discovered, but it is difficult to dig out the significant rules with high prevalence and low support at the same time, resulting in great uncertainty and inaccuracy in the MDR patterns. To solve this problem, some studies discarded the threshold support and confidence and used the fuzzy algorithm to discover more rules (Bansal et al., 2017). However, the results are miscellaneous and relatively poor, and it is difficult to accurately locate the required association rules.

In order to dig out effective and scientific rules, we believe that a reasonable weighted threshold value must be established for rule identification. Therefore, in this study, the drug resistance prevalence of bacteria was considered as the weight values, then a weighted Apriori association rule algorithm was proposed to determine the multidirectional relationship between drug resistance of different antibiotics caused and explore the MDR patterns of bacteria in cow mastitis.

Finally, to compare the relative MDR rules and confirm the high correlation between of different bacteria, a method is proposed for comparing MDR based on the Euclidean metric similarity algorithm using Staphylococcus aureus (S. aureus) and coagulase negative staphylococci (CNS) as the study subjects from the overall level with support, confidence, and lift as characteristic parameters. This study aims to guide the rational use of antibiotics in the farming industry and provide a scientific basis for relevant departments to deal with drug resistance.

Section snippets

Data collection

Pathogenic bacteria were isolated and identified from the samples of cows with mastitis in large-scale farms, and then drug sensitivity tests were conducted. A total of 13,425 pieces of drug resistance data was collected. The data content includes the ID, site, time, quantity, minimum inhibitory concentration (MIC), and drug resistance of the bacteria, providing a complete data basis for the MDR analysis from the perspective of data-driven.

After statistical analysis, the bacteria of mastitis in

Overall results and visualization

In this section, S. aureus, CNS are used as examples for their MDR analysis. All possible interdependencies are considered for weighted frequent itemset mining and generating association rules based on the proposed algorithm and MDR dataset.

The generation rules are shown in Fig. 1. A total of 1442 rules were generated for S. aureus, 616 rules for CNS, indicating that MDR in S. aureus is more serious and the drugs are mainly used for S. aureus, less for CNS. This also shows that the drug

Conclusions

In this paper, a weighted association rule mining algorithm was proposed to determine the MDR patterns of bacteria isolated from cow mastitis. Then a method was proposed for comparing the similarity of MDR rules in different bacteria. The results showed that high drug resistance to penicillin, gentamicin, and ampicillin, and most other antibiotics were linked with these three antibiotics. Besides, although the main rules are pretty similar in pathogenic bacteria, the MDR rules in S. aureus are

CRediT authorship contribution statement

Buwen Liang: Conceptualization, Methodology, Software, Formal analysis, Writing – original draft, Writing – review & editing. Xinxing Li: Conceptualization, Validation, Formal analysis, Resources, Writing – review & editing, Visualization. Ziyi Zhang: Software, Formal analysis. Congming Wu: Validation, Investigation, Resources, Data curation, Project administration, Funding acquisition. Xin Liu: . Yongjun Zheng: Conceptualization, Validation, Investigation, Resources, Data curation,

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This study was supported by the China Agriculture Research System of MOF and MARA (grant No. CARS-36) and National Key Research and Development Project of China (grant No. 2016YFD0501304).

Reference (35)

  • P. Collignon et al.

    Anthropological and socioeconomic factors contributing to global antimicrobial resistance: a univariate and multivariable analysis

    Lancet Planet. Heal.

    (2018)
  • Acan, H., 2017. On a uniformly random chord diagram and its intersection graph. Discrete Math. 340, 1967–1985....
  • R. Agrawal et al.

    Fast Algorithms for Mining Association Rules

  • Bansal, M., Grover, D., Sharma, D., 2017. Sensitivity Association Rule Mining using Weight based Fuzzy Logic. Glob. J....
  • Cazer, C.L., Al-Mamun, M.A., Kaniyamattam, K., Love, W.J., Booth, J.G., Lanzas, C., Gröhn, Y.T., 2019. Shared Multidrug...
  • Ceglar, A., Roddick, J.F., 2006. Association mining. ACM Comput. Surv....
  • W.N. Cheng et al.

    Bovine mastitis: risk factors, therapeutic strategies, and alternative treatments — A review. Asian-Australasian

    J. Anim. Sci.

    (2020)
  • K. Clifford et al.

    Antimicrobial resistance in livestock and poor quality veterinary medicines

    Bull. World Health Organ.

    (2018)
  • R. Davies et al.

    Antimicrobial Resistance on Farms: A Review Including Biosecurity and the Potential Role of Disinfectants in Resistance Selection

    Compr. Rev. Food Sci. Food Saf.

    (2019)
  • Gao, X., Fan, C., Zhang, Z., Li, S., Xu, C., Zhao, Y., Han, L., Zhang, D., Liu, M., 2019. Enterococcal isolates from...
  • Hipp, J., Güntzer, U., Nakhaeizadeh, G., 2000. Algorithms for association rule mining — a general survey and...
  • Hudson, J.A., Frewer, L.J., Jones, G., Brereton, P.A., Whittingham, M.J., Stewart, G., 2017. The agri-food chain and...
  • Hvistendahl, M., 2012. China takes aim at rampant antibiotic resistance. Science (80-.). 336, 795....
  • T.A. Kumbhare et al.

    An Overview of Association Rule Mining Algorithms

    (2014)
  • Lamma, E., Riguzzi, F., Storari, S., Mello, P., Nanetti, A., 2003. Discovering validation rules from microbiological...
  • X. Li et al.

    Antimicrobial Resistance Risk Assessment Models and Database System for Animal-Derived Pathogens

    Antibiotics

    (2020)
  • Ludwig, A., Berthiaume, P., Boerlin, P., Gow, S., Léger, D., Lewis, F.I., 2013. Identifying associations in Escherichia...
  • Cited by (3)

    View full text