Development of a predictive statistical model on ARG abundance in sediment sediments

Abstract

Antibiotics are important substances for humans and animals to destroy or limit the growth of disease disease- causing bacteria bacteria. The use of antibiotics globally was increasing every year with a 35 % increase in the years 2000 t o 2010 2010. However, antibiotic resistance is a growing problem worldwide causing 700.000 deaths per year . The capabilities of microbes in exchanging antibiotic resistant genes (ARGs) and use of antibiotics by mankind, led to the increasing antimicrobial resistance worldwide. The presence of antibiotics in the environment has been shown to correlate to ARGs and differ per type of
compartment, such as sediment sediment. Thereforeherefore, antibiotics have been sugges ted to exert selective pressure on bacteria . Since there is lack of biological mechanistic understanding, statistical modelling might help
to understand the relationship between antibiotic concentration s, ARGs and influencing factors in sediments.
In this thesis a statistical model is developed based on literature data, which explains resistant genes with antibiotic selective pressure in sediment. Literature research has been done to collect data on ARGs and antibiotics measured in sediment. After a process of selection on the usability of the data, 22 studies were used to supply the data. Total relative abundance of resistant genes (TARG) and total selective pressure per antibiotic class (TASP) were calculated. TARG and TASP were linked to each other by antibiotic class class, which resulted in a total of 1122 data points. Data exploration was conducted to examine the relations within the data set and a full linear mixed effectffects model was developed to explain the dataset with a statistical model . Within the full model, TARG is the response variable and TASP is the explanatory variable. Antibiotic class ( Class ) and environmental compartment Compartment ) were added as categorial explanatory variables. Study, Year, Sample, Country were added as covariates . Term (Compartment| and nested term (1|Study/ completed the full model. To find the most parsimonious model among the 35 created po tential models, evaluation of the random structure and the fixed structure was done.
The data exploration showed no correlation between TARG and TASP . However, when splitting the data into subsets according to a factor factor, positive correlations were observed observed. The TARG and TASP subset for lakes and the subset for class sulfonamides in rivers showed significant correlations.
Additionally, a strong positive correlation was observed between macrolides and resistant gene qnr( A). After assessment of the model models with the best fit for the data , TASP was not an explanatory factor for the dataset. The most significant fixed factors were Compartment and Class and t he most significant
random factors were (1|Study/ and (1 / Sample). The full model explained 86% of the variance in the data. TASP was not found to be an explanatory factor for TARG TARG, which shows that the association between TARG and TASP is complex, non non-linear and possible other fa ctors can contribute to the
abundance and sprea d of ARGs.

Date of Award	19 Dec 2023
Original language	English
Supervisor	Jikke van Wijnen (Examiner), Ad Ragas (Supervisor) & Lily Fredrix (Assessor)