Temporal Dynamics of Microbial Communities in Anaerobic Digestion: Influence of Temperature and Feedstock Composition on Reactor Performance and Stability

Temporal Dynamics of Microbial Communities in Anaerobic Digestion: Influence of Temperature and Feedstock Composition on Reactor Performance and Stability | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results Temporal Dynamics of Microbial Communities in Anaerobic Digestion: Influence of Temperature and Feedstock Composition on Reactor Performance and Stability View ORCID Profile Ellen Piercy , Xinyang Sun , View ORCID Profile Peter R. Ellis , Mark Taylor , View ORCID Profile Miao Guo doi: https://doi.org/10.1101/2025.01.12.632589 Ellen Piercy a Department of Engineering, Faculty of Natural, Mathematical & Engineering Sciences, King’s College London , Strand Campus, London, WC2R 2LS, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Ellen Piercy Xinyang Sun a Department of Engineering, Faculty of Natural, Mathematical & Engineering Sciences, King’s College London , Strand Campus, London, WC2R 2LS, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Peter R. Ellis b Biopolymers Group, Departments of Biochemistry and Nutritional Sciences, Faculty of Life Sciences & Medicine, King’s College London , Franklin-Wilkins Building, London, SE1 9NH, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Peter R. Ellis Mark Taylor c Fermentation Lead, Marlow Ingredients , Nelson Ave, Billingham, North Yorkshire, TS23 4HA, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Miao Guo a Department of Engineering, Faculty of Natural, Mathematical & Engineering Sciences, King’s College London , Strand Campus, London, WC2R 2LS, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Miao Guo For correspondence: miao.guo{at}kcl.ac.uk Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF Abstract Anaerobic digestion (AD) offers a sustainable biotechnology to recover resources from carbon- and nutrient-rich wastewater streams, such as food-processing wastewater. Despite the wide adoption of crude wastewater characterisation, the impact of detailed chemical fingerprinting on AD remains underexplored. This study investigated the influence of fermentation-wastewater composition and operational parameters on AD over time to identify critical parameters influencing microbiome diversity and reactor performance. Eighteen bioreactors were operated under various operational conditions using mycoprotein fermentation wastewater. Detailed chemical analysis fingerprinted the molecules in the fermentation-wastewater throughout the AD process including sugars, sugar alcohols and volatile fatty acids (VFAs). High-throughput sequencing revealed distinct microbiome profiles linked to temperature and reactor configuration, with mesophilic conditions supporting a more diverse and densely connected microbiome. Importantly, significant elevations in Methanomassiliicoccus were correlated to high butyric acid concentrations and decreased biogas production, further elucidating the role of this newly discovered methanogen in AD. Reactors from different experimental runs had distinct VFA profiles, which impacted microbial taxonomy and diversity. Dissimilarity analysis demonstrated the importance of individual VFAs, sugars and sugar alcohols on microbiome diversity, highlighting the need for detailed chemical fingerprinting in AD studies of microbial trends. Furthermore, machine learning models predicting reactor performance achieved high accuracy based on operational parameters and microbial taxonomy. Operational parameters were found to have the most substantial influence on chemical oxygen demand removal, whilst Oscillibacter and two Clostridium species were highlighted as key factors in biogas production. By integrating detailed chemical and biological fingerprinting with explainable machine learning models this research presents a novel approach to advance our understanding of AD microbial ecology, offering insights for industrial applications of sustainable waste-to-energy systems. 1. Introduction Global wastewater is estimated to have reached 400 billion m 3 /year 1 . Alarmingly, 48% of wastewater is inadequately treated, posing significant environmental and public health risks 2 . Compared with municipal wastewater, food-processing wastewater is often characterised by high organic matter concentrations, which require advanced treatment technologies 3 . Aerobic wastewater treatments are energy-intensive, cost ineffective, and contribute to greenhouse gas emissions 4 . This realisation has prompted a shift towards more sustainable approaches, which recover valuable compounds from wastewater. Anaerobic digestion (AD) offers a biological process that converts organic matter to energy-carrier gas (biogas and nutrient-rich digestate), which has the potential to reduce environmental impacts and operational costs 5 , 6 . Many studies have focused on lab-scale reactors using synthetic feedstock, such as acetic acid or glucose-based growth mediums ( Figure 1a ). However, synthetic feedstock is not representative of the fluctuations and contaminants within real-world wastewater. Detailed chemical composition of wastewater feedstock remains largely undefined, especially within studies including detailed biological characterisation. Diverse microbes show different substrate affinities; therefore, it is important to understand how the molecules present in wastewater streams impact microbial diversity ( Figure 1b ). Moreover, previous studies often focus on singular time points, which do not capture the dynamic nature of the microbiome in response to fluctuating feedstock structure and composition. Understanding the interactions between these parameters and microbial communities over temporal scales is essential for enhancing biogas production, process stability, and overall efficiency. Download figure Open in new tab Download figure Open in new tab Figure 1. Growth conditions reported to support microbial growth for anaerobic digestion from a literature review of 197 papers (supplementary materials and methods 1.1). (a) Distribution of feedstock, reactor configuration and operational temperature. (b) Substrates (187) reported to support microbial growth were investigated. Affinity was based on average substrate utilisation on a scale of 0 to 1, where 1 (green) indicates that substrate supports the growth of all species within a phyla, and 0 (black) indicates no reported growth on a given substrate. Substrates are categorised by key functional groups (Supplementary Table 1.1). (c) Pie charts displaying the number of genera belonging to 16 bacteria phyla reported to utilise different categories of substrate. (d) Temperature and pH range reported to support growth. Temperature is categorised as psychrophilic ( temperature >42°C) or thermophilic (>42°C). The pH range is categorised as acidic (≤ 6.5, pink), neutral (6.5< pH <7.5, light green), or alkaliphilic (≥7.5, dark green). NA (grey) indicates information not available; n represents number of genera. The findings of a comprehensive literature review comprising 197 publications of AD species growth and experimental conditions are synthesised in Figure 1 . On the phylum level, Firmicutes has the most diverse substrate utilisation, which could help to explain the wide diversity of Firmicutes reported in AD ( Figure 1b ). The majority of Firmicutes and Bacteroidetes species utilised sugars, sugar alcohols, polysaccharides and carboxylates. Specific substrate categorisation is detailed in according to Supplementary Table 1.1. Conversely, Proteobacteria showed complementary substrate utilisation to Bacteroidetes , with higher utilisation of amino acids, proteins, redox substrates (such as hydrogen and nitrates) and pyridines, benzoates and phenols ( Figure 1c ). These findings provide insights into how different feedstocks influence microbial community identity. However, the results are not comprehensive, and further research is required to relate detailed feedstock chemical characterisation with various microbial enrichments. Operational parameters play a crucial role in AD performance and stability. In industrial-scale AD plants, microbial communities are subjected to fluctuating conditions, including feedstock structure and composition, temperature, and pH variations 7 , 8 . These fluctuations could lead to community changes, potentially causing process instability 9 . Most species had a mesophilic optimum temperature (63%); however, many species had a wide range of temperatures that could support growth ( Figure 1d ). Most bacteria (56%) had an optimum pH range at or close to neutral, although one-quarter of species did not have a specified pH growth range ( Figure 1d ). Archaea had a higher proportion of alkaliphilic species compared to more acidophilic bacteria (21% versus 11%, respectively, Figure 1d ). Furthermore, machine learning (ML) is a rapidly expanding field that is being applied to human microbiome studies 10 , 11 . Nevertheless, the application of ML in applied and industrial microbiology remains an emerging field. Current studies have shown promise for predicting AD performance from operational and biological parameters 10 . However, these studies suffer from limited sample sizing (between 17 to 50 samples), potentially limiting statistical power, and also lack detailed feedstock and biological characterisation 11 , 12 . Furthermore, model interpretability is often limited due to the need to reduce high dimensionality datasets. Applying explainable ML techniques, such as Shapley Additive exPlanations (SHAP) values, to datasets including detailed chemical and biological fingerprinting, allows interpretability of factor impact to improve our understanding of how operational, chemical and biological factors contribute to AD reactor performance. To address this knowledge gap, our research aimed to understand the resource recovery potential of complex wastewater streams from fermentation processes and investigate the dynamics of the microbiome underpinning AD across different operational parameters and temporal scales. Here, we present the results of experiments exploring the effect of feedstock composition on AD using characterised mycoprotein fermentation wastewater (MFWW) obtained from the food industry. 2. Materials and methods MFWW, a by-product of Quorn™ food production, was obtained from Marlow Ingredients (Billingham, UK). Four batches of MFWW were characterised to evaluate batch variation and the effect of the feedstock on AD ( Figure 2a ). Two experiments were designed to address the research objectives. In the first experimental run, mesophilic and thermophilic reactors were compared to evaluate the effect of temperature. In run 2, single- and two-stage reactors were assessed to determine the impact of reactor configuration ( Figure 2b-c ). Influent composition was monitored in a time-series sampling schedule ( Figure 2d ). Reactor chemistry, performance and microbiome biodiversity were analysed to address how MFWW composition affects these parameters across different operational parameters and temporal scales. Predictive ML models were developed to forecast AD reactor performance in terms of chemical oxygen demand (COD) removal, cumulative biogas and specific daily biogas based on biological, chemical and operational parameters. Download figure Open in new tab Figure 2. Experimental design. (a) Mycoprotein fermentation wastewater characterisation. (b) Experimental grouping. (c) Bioreactor set up, sampling and analysis. (d) Sampling schedule. *Abbreviations: chemical oxygen demand (COD), total nitrogen (TN), biomethane potential (BMP), high performance liquid chromatography (HPLC), volatile fatty acid (VFA), and continuously stirred tank reactor (CSTR). 2.1. Digester description Eighteen 5.5L continuously-stirred tank reactors (CSTR) were operated with a 14.7-day hydraulic retention time and 0.32L/d flow rate with MFWW ( Figure 2c ). Reactors were operated for a two-month adaptation period with MFWW as feedstock at 0.18L/d. CSTR were grouped by operational conditions. Group 1 represented single-stage mesophilic reactors (37±1°C), group 2 represented single-stage thermophilic reactors (50±1°C) and group 3-5 represented mesophilic single-, first- and second-stage reactors, respectively ( Figure 2b ). Groups 1 and 2 contained biological triplicates, and groups 3, 4 and 5 contained four biological replicates each. 2mL samples were taken in a time-series schedule in triplicate, and operational parameters were monitored throughout ( Figure 2d ). Samples were evaluated using the methods detailed below. Groups 1-2 and 3-5 were set up as two independent runs using the same batch of inoculum obtained from a commercial food-waste AD plant treating municipal source-separated domestic kitchen waste, industrial food waste and a fraction of abattoir waste (Devon, UK). Volatile suspended solid content was quantified as a measure of active biomass using standard methods (Supplementary Figure 2.1) 13 . 2.2. Analytical methods 2.2.1. Biogas production Real-time biogas production and composition (CH 4 and CO 2 ) from each individual reactor was measured in real-time using an Anaero technology gas flow meter and sensor. 2.2.2. Chemical oxygen demand (COD) COD was measured directly from MFWW and reactor samples using Vario HR/COD, MR/COD, LR/COD Lovibond kits (as appropriate), according to standard methods (ISO, 1989; Supplementary materials and methods 2.3). COD removal was calculated using Supplementary Equation S3. 2.2.3. pH Fresh MFWW and reactor samples were measured directly in triplicate with a pH meter (Mettler Toledo). 2.2.4. High performance liquid chromatography Analysis of sugars and sugar alcohols (melibiose, glucose, maltitol, glycerol, mannitol and arabitol) and VFAs (acetic, propionic, isobutyric, butyric, isovaleric and isovaleric acid) for MFWW and reactor samples were conducted using a Shimadzu U-HPLC with SPD-M40 photodiode array and RID-20A refractive index detectors (Shimadzu). Prior to analysis samples were filtered through a 0.22µm pore filter (Claristep® Filtration system, Sartorius). Analysis was carried out in triplicate, and the mean and standard deviation were calculated. Calibration curves were constructed by plotting peak area against concentration. The linearity of the line of best fit was calculated using the least square regression method within the LabSolutions software. The limit of detection and quantification were calculated using LabSolutions software to verify quality. Detailed methodologies are provided in Supplementary Materials 2.4. 2.3. Metagenomic Sequencing and Bioinformatics 2.3.1. DNA extraction Reactor samples were centrifuged at 10,000rpm for 5 minutes to obtain a pellet which was used for the DNA extraction, using the DNeasy® PowerSoil® Pro Kit according to standard protocol (QIAGEN GmbH, Germany). PCR amplification was carried out using 16S rRNA primers (Supplementary Figure 2.5.1c). Amplicon barcoding PCR was conducted using BiomekFX robot and automated software. Sequencing was conducted using Illumina MiSeq v3 kit and the run quality was assessed based on quality score (% bases>Q30), raw cluster density (k/mm 2 ), data output (Gbp), read number and % PhiX aligned (Supplementary Figure 2.5.1). 2.3.2. Sequence analysis Taxonomic classification was conducted using the Mothur pipeline in Galaxy referenced against the Silva_v4 database with ≥97% sequence identity threshold 14 , 15 . Additional pre-processing steps were conducted to filter for sequences ≥1200bp with Phred qualities >20 per base (Supplementary Figure 2.5.2). 2.3.3. Phylogenetic tree construction A Newick tree of genera was constructed from the taxonomic classification data using the anytree library in python and visualised using the international tree of life 16 . 2.3.4. Network analysis The relative abundance of classified microbial taxa for each reactor sample were analysed across all time points, using the Molecular Ecology Network Analysis pipeline (Supplementary Materials and Methods 2.6) 17 . Networks were visualised in Cytoscape at the genus level 18 . 2.3.5. Alpha diversity Shannon, Simpson’s and Chao1 diversity indices were calculated using Supplementary Equation S4-6. 2.3.6. Beta diversity and dissimilarity index The calculation of Bray-Curtis similarity and db-RDA was implemented using the scipy.spatial, skbio and statsmodels library in python (Supplementary Eq. S7) 19 . The dissimilarity matrix served as the response variable, while environmental variables were used as predictors. Detailed methodology is available in Supplementary Materials 2.8. 2.4. Machine learning models Lasso regression, random forest and bagging regression models were evaluated using algorithms defined in Eq.S9-11 (Supplementary Materials 2.9). The models were evaluated based on their performance in predicting reactor performances. Root mean squared error (RMSE), and coefficient of determination (R²) were used to assess performance. 2.5. Statistical analysis For normally distributed data with two comparisons a pairwise t-test was applied, for three or more comparisons a one-way ANOVA analysis was performed with Tukey post-hoc analysis. For non-normally distributed independent data a Kruskal-Wallis H test was conducted with Dunn’s post-hoc test for significant results. Confidence intervals of 95%, 99% and 99.9% were used. To investigate correlations between syntrophic relationships, a non-parametric Spearman’s rank correlation analysis was conducted. 3. Results 3.1. MFWW represents a complex feedstock suitable for AD The composition of four different batches (A-D) of MFWW, compared with a previously reported value (sample R) obtained from the food industry, were measured over the time course of four fermentation cycles to determine resource recovery potential ( Figure 3 ). We quantified physicochemical parameters driving the AD process, including ammonium, protein, pH, COD, and soluble organic compounds in MFWW ( Figure 3 ). MFWW presented as a complex but balanced feedstock with a neutral pH (6.2±0.1) and high potential for AD ( Figure 3a ). Inhibitory ammonium was present at low concentrations (0.03±0.01g/L, Figure 3b ). High COD (11.32±4.09g/L) and a mixed sugar profile, paired with high protein content (2.19±1.31g/L) facilitating good C/N ratios and suggests high resource recovery potential ( Figure 3c-e ). Biomethane potential assays showed high biodegradability (92±6%, Figure 3f ). Download figure Open in new tab Figure 3. Chemical characterisation for four different batches (A-D) of mycoprotein fermentation wastewater (MFWW) compared to previously internally reported values (batch R). Batch C was measured at 14 times points within a 30-day fermentation cycle. Violin plots represent the distribution of results, box plots represent the interquartile range, the central bar represents the mean average, error bars represent the standard deviation and dots represent outliers. Statistical significance compared to R are shown at 95% *, 99% ** and 99.9% *** confidence intervals 20 , 21 . (a) pH. (b) Ammonium. (c) Chemical oxygen demand. (d) Protein. (e) Sugars and sugar alcohols. (f) Biomethane potential assay results including cumulative biogas of MFWW compared to inoculum only, acetic acid, propionic and butyric acid, and glucose. Final biomethane production compared as a percentage of the theoretical value calculated based on the COD to CH 4 mass balance equation. Error bars represent standard deviation. 3.2. Bioreactor performance 3.2.1. Run 1 reactors had better COD removal but poorer biogas production than Run 2 reactors COD removal in samples obtained from run 1 reactors were consistently greater than the 75% regulatory threshold and run 2 reactor samples (p=0.000, Figure 4b ) 22 . COD removal efficiency in run 2 reactor samples regularly fell below the regulatory threshold, particularly at the beginning of the experiment up to day 28 and again at day 38, but gradually increased over time ( Figure 4b ). Conversely, run 2 reactors produced higher cumulative biogas than run 1 (p=0.000, Figure 4b ). We compared the chemical and biological differences between reactors with high and low biogas production ( Figure 4 - 5 ). Download figure Open in new tab Figure 4. Reactor operation, performance and chemistry (a) Influent composition including mean (Avg.) ± standard deviation (SD) for pH, chemical oxygen demand (COD), total solids (TS), volatile solids (VS), total suspended solids (TSS), volatile suspended solids (VSS), ammonium, nitrites, nitrates, total nitrogen (TN), protein, sulphate, phosphate, total sugar & sugar alcohol, melibiose, maltitol, glucose, glycerol, mannitol, arabitol, total volatile fatty acid (VFA), formic acid, acetic acid, propionic acid, isobutyric acid, butyric acid, isovaleric acid and valeric acid. All units are g/L (except pH). ( b) Performance parameters including COD removal, cumulative biogas, and specific daily biogas. ( c) Reactor pH and VFA content. Bar charts represent the cumulative sum of individual VFA over time (days). Line chart represents pH (grey). Mean values ± standard deviation are presented for VFA concentrations and pH. Download figure Open in new tab Download figure Open in new tab Figure 5. CSTR taxonomy (a) Phylum level relative abundance (% of total population), phyla with >1% relative abundance for at least one sample are represented and phyla representing 1% relative abundance for at least one sample are represented and taxa representing <1% of the total population for any sample are aggregated into ‘other’. Taxa are represented at the kingdom (_k), phylum (_p), class (_c), order (_o), family (_f) and genus (_g) level. (c) Relative abundance of archaea categorised by methanogenic pathway, where other represents ambiguous taxa. (d) Alpha diversity for Shannon, Simpson’s and Chao1 diversity indices over time, days. (e) Core microbiome of species shared across all reactors and unique species from experimental run 1 and 2. Archaea are highlighted in green, Actinobacteria in yellow, Bacteroidetes in red, Firmicutes in blue, Proteobacteria in orange, and other bacteria are light purple. Mesophilic reactors had better biogas production than thermophilic (p=0.000) despite similar COD removal (p=1.000, Figure 4b ). This result indicates that thermophilic conditions created an unfavourable environment for biogas production, possible due to poor adaptation increased temperature ( Figure 4b ). This could be indicative of the inoculum coming from a mesophilic source, meaning that thermophilic genera were underrepresented. Another possibility is that the temperature was low enough to support mesophilic genera with thermotolerant traits, but caused thermal stress, leading to underperformance, suggesting a specialised microbiome with little flexibility for adaptation ( Figure 4b ). 3.2.2. VFA content showed marked differences between Run 1 and 2 reactor samples Samples obtained from run 2 had elevated VFA concentration compared to run 1 (p=0.000) and were characterised by higher acetic (p=0.000), isovaleric (p=0.020) and valeric acid (p=0.000, Figure 4c ). Isobutyric acid constituted a large proportion of the VFA content produced for all run 1 reactors, likely owing to the influence of the influent containing amino acids, such as valine and leucine, which can be converted during AD to isobutyric acid and isovaleric acid, respectively 23 – 25 . The influent contained relatively high concentrations of isobutyric acid and isovaleric acid ( Figure 4a ) 26 . CSTR3.1-3.2, had elevated total VFA contents compared to CSTR3.3-3.4 (p=0.007 and p=0.001, respectively, Figure 4c ). CSTR3.1 had elevated isovaleric acid, and CSTR3.2 had significantly elevated valeric acid compared to other group 3 reactors (p=0.000, Figure 4c ). Samples in run 1 and 2 reactors had a spike in VFA content on day 7 and 13, respectively ( Figure 4c ). This peak consisted of elevated isovaleric content in samples for most reactors (except CSTR3.1) and increased acetic acid for run 2 reactors ( Figure 4c ). For second-stage reactors CSTR5.2, had elevated valeric acid (p=0.000), whilst CSTR5.1 had lower total VFA content than other group 5 reactors (p<0.05, Figure 4c ). The average reactor pH ranged from 8.00 to 8.44 and was slightly higher for groups 1, 2 and 5 than 3 and 4 (p=0.000, Figure 4c ). Group 2 reactors pH showed high variability, suggesting potential instability ( Figure 4c ). Reactor pH is a key factor influencing microbial activity, and different pH ranges support different species’ growth ( Figure 1d ). Therefore, it will be interesting to investigate the effect of this pH difference by comparing microbial communities between reactors ( Figure 1d , Figure 4c ). 3.3. Microbiome composition and function 3.3.1. Run 1 reactor samples had a distinct taxonomic profile, characterised by increased Methanosaeta compared to run 2 samples In total, 1,982,343 unique sequences were retained after quality control over 25 times points. Archaea accounted for 1.23±0.72% of the total population. Proteobacteria, Firmicutes, Bacteroidetes, and Actinobacteria represented the major phyla across all reactors, although run 1 had a distinct taxonomic profile compared to run 2 reactors ( Figure 5a ). Notably, the relative abundance of Firmicutes was significantly elevated in run 1 compared to run 2 reactors, indicating elevated Firmicutes is associated with reduced biogas production (p<0.05, Figure 5a ). Additionally, the relative abundance of Synergistota was significantly elevated in group 1 reactors compared to groups 2, 3 and 4 ( p =0.000, Figure 5a ). 3.3.2. The Firmicutes taxonomic profile was highly dynamic Core microbiome construction and times series network analysis revealed that Firmicutes was highly dynamic, with little conservation over time for run 2 reactors with high biogas production ( Figure 4b , Figure 5e ). This result suggests that a dynamic Firmicutes community improves reactor performance ( Figure 5e ). Interestingly, Firmicutes had the most diverse substrate utilisation of key AD phyla, which could likely be contributing to their dynamic profile ( Figure 1b ). By contrast, there was a much higher presence of conserved Proteobacteria and Bacteroidetes across run 2 reactors that were not shared by run 1 ( Figure 5e ). 3.3.3. Taxonomic profile appeared to be decoupled from reactor performance Despite having the same feedstock, operational conditions and starting inoculum, CSTR3.1 had a distinct taxonomic profile compared to other run 2 reactors, but comparable biogas production and COD removal ( Figure 4b , Figure 5 ). Run 1 reactors had a higher abundance of Methanosaeta ( p =0.000), corresponding to lower acetic acid concentration, suggesting active acetoclastic methanogenesis ( Figure 5b ). Methanosaeta is an obligate acetoclastic methanogen associated with low organic loading rate (OLR), acetic acid, and solid and ammonium contents 27 . Additionally, Methanosaeta are reported to have an alkaliphilic optimum pH of 7.0 to 9.0 ( Figure 1d ). The higher pH and lower acetic acid concentration of run 1 compared to group 3 and 4 could explain the elevated Methanosaeta ( Figure 4c , Figure 5b ). Notably, CSTR3.1 had similar performance to run 2 reactors but a taxonomic profile that was more comparable to run 1; for example, CSTR3.1 and run 1 reactors had comparable Methanosaeta abundance ( Figure 5 ). This result indicates that reactor performance was decoupled from the microbiome profile. Despite low OLR, acetic acid concentration and high pH, Methanosaeta content was reduced in group 5 reactors, indicating potentially less reliance on acetoclastic methanogenesis ( Figure 5b-c ). However, second-stage reactors had a lower percentage of classified genera compared to first- and single-stage reactors making it difficult to conclude the true relative abundance of acetoclastic methanogens ( Figure 5 ). 3.3.4. Despite some fluctuations, reactors had similar and stable species richness and evenness Shannon, Simpson’s and Chao1 alpha diversity indices were compared as measures of community evenness and richness ( Figure 5d ). There was a significant drop in Shannon and Simpson’s diversity for only CSTR3.3 on day 43, which recovered by day 48 ( Figure 5d ). By day 28, Simpson’s diversity was similar for all reactors ( Figure 5d ). Chao1 diversity was similar for CSTR1.1 and Group 3; however, CSTR1.2 and CSTR1.3 showed large fluctuations, resulting in periods of elevated Chao1 diversity index values ( Figure 5d ). Reactor CSTR1.2 had elevated Chao1 diversity past day 20, whereas CSTR1.3 had elevated Chao1 diversity between days 1 and 15, indicating more rare species than other reactors ( Figure 5d ). Species richness and evenness were reduced in group 5 reactors compared to group 1-4 reactors, indicating that alpha diversity plays a crucial role in reactor resilience to stress conditions. This result agrees with the literature that diversity improves performance ( Figure 5d ) 28 . 3.4. Microbiome dynamics and interaction with bioreactor parameters 3.4.1. Denser networks were associated with increased biogas production Run 1 reactor samples were found to represent less dense networks than those in run 2 ( Figure 6a ). These results suggest that network density contributes to reactor performance, supporting previous findings from Dalantai et al., 29 , who reported that sparser networks were associated with reactor instability. Time series analysis of node centrality showed that groups 3 and 5 had a higher node degree than groups 1, 2 and 4 ( Figure 6b ). Group 4 showed greater variation in node betweenness and a lower clustering coefficient than other reactor groups ( Figure 6a ). Interestingly, run 1 had a higher node event centrality than run 2, suggesting that node event centrality could be important in reactor performance (p= 0.000, Figure 6b ). Download figure Open in new tab Download figure Open in new tab Figure 6. Microbiome interaction analysis. (a) Time series network analysis of each reactor group at genus level, where nodes represent genera. (b) Node centrality is presented where bar charts represent the average, and error bars represent the standard deviation, and line charts represent node centrality over time for node degree, betweenness, clustering coefficient and event centrality. (c) Descriptive statistics of network analysis and the identity of nodes with maximum degree, betweenness and centrality are described. (d) Spearman rank correlation coefficient of dominant methanogens and syntrophic genera presented as a heatmap for group 3-5 reactors, where c orrelations are reported as positive (red) or negative (blue). Blank values indicate insufficient data. Only correlations with sufficient data for at least one genera are presented. 3.4.2. Syntrophic relationships were associated with increased biogas production Syntrophic bacteria and methanogens, including Syntrophomonas, Methanosarcina and Methanosaeta , represented group 1 central network nodes ( Figure 6c ). CSTR2.3 also had high connectivity of syntrophic genera ( Figure 6c ) and elevated biogas production compared to other group 2 reactors, indicating the potential importance of syntrophic genera connectivity in reactor performance ( Figure 4b , Figure 6c ). Mesophilic reactors had elevated syntrophic genera compared to thermophilic reactors, and higher connectivity of syntrophic genera and methanogens was associated with improved biogas production. Syntrophic genera belonging to Firmicutes , which have been related to stress conditions, such as high OLR, were present across all run 1 reactors, which could have contributed to lower biogas production 30 , 31 . Spearman’s rank correlation analysis revealed significant correlations between thermophilic Caloramator and Methanosaeta and Methanosarcina in thermophilic reactors only ( Figure 1 , Figure 6d ). Spearman rank correlation analysis revealed a positive correlation between hydrogen-producing syntrophic bacteria and hydrogenotrophic methanogens in single- (group 3) and first-stage (group 4) reactors, which was diminished in the second-stage (group 5) reactors ( Figure 6d ). Most syntrophic bacteria showed significant correlations to dominant methanogens for single- and first-stage reactors (p0.05, Figure 6d ). However, the positive correlation between Candidatus Cloacamonas and Methanomassiliicoccus was significant for second-stage reactors (p=0.000, Figure 6d ). Anaerolinea and acetoclastic methanogens were significantly positively correlated; however, the strength of this correlation was lower in single-compared to first- and second-stage reactors (p<0.05, Figure 6d ). The relationship between syntrophic genera were less clear in samples from second-stage reactors, which were found to have fewer significant correlations ( Figure 6d ). However, Anaerolinea, Methanosarcina, and Methanosaeta showed significant positive correlations (p=0.000, Figure 6d ). Desulfuromonas and Pseudomonas were also significantly positively correlated to Methanoculleus, Methanosaeta and Methanomassiliicoccus and Geobacter was significantly correlated with Methanoculleus (p=0.000, Figure 6d ). Importantly, samples in the group 5 reactors exhibited decreased acetic acid concentrations and acetoclastic methanogenesis, which could reduce the strength of syntrophic acetate oxidising bacteria relationships within the reactors ( Figure 4c , Figure 6d ). This result highlights the importance of syntrophic relationships in reactor performance. 3.4.3. Methanomassiliicoccus was associated with increased butyric acid concentration and decreased biogas production Group 2 reactors had a greater abundance of Methanomassiliicoccus than other reactors (p=0.000, Figure 5b ). Interestingly, Methanomassiliicoccus were associated with periods of enriched butyric acid, suggesting a potential correlation between Methanomassiliicoccus and isobutyric and butyric acid concentrations ( Figure 4c , Figure 5b ). This result is consistent with data from Nikitina et al., 32 , who found that Methanomassiliicoccus were enriched under butyric acid dominant VFA accumulation. Methanomassiliicoccus has been associated with reduced biogas yields, which agrees with the results that run 1 reactors had lower biogas production than run 2 reactors ( Figure 4c , Figure 5 ) 33 . 3.5. Effects of chemical and operational parameters on bioreactor performance and biodiversity 3.5.1. Random forest models performed best for predicting anaerobic digestion reactor performance Supervised ML models were developed to investigate the potential of operational parameters and reactor chemistry and biology for predicting reactor performance in terms of COD removal and biogas production. Lasso regression models provided a baseline with moderate predictive capability. Random forest regression is a robust ensemble method that creates and merges multiple decision trees for a more accurate and stable prediction. Bagging regression is an ensemble technique that improves the stability and accuracy of ML algorithms by reducing variance and preventing overfitting. Random forest regression showed the best performance with the highest RMSE and lowest R², reflecting high accuracy and substantial variance explanation ( Figure 8b-d ). Overall, the evaluation revealed that the random forest outperformed the two other models across all experiments ( Figure 8b-d ). Download figure Open in new tab Figure 8. Effect of operational and chemical parameters on microbiome biodiversity and performance. (a) Dissimilarity analysis of mesophilic (M1 to M3, purple) and thermophilic (T1 to T3, pink) reactors assessing the influence of operational parameters, influent VFA content, influent sugar and sugar alcohol content, and reactor VFA content. Machine learning predictive models includes Lasso regression, random forests and bagging regression. The inputs and outputs of each model are described in the tables, the RMSE and R 2 values used to evaluate the models are presented as bar charts and factor influence is displayed in terms of SHAP values for each model predicting reactor performance in terms of (b) COD removal. (c) Cumulative biogas, L. (d) Specific daily biogas, L biogas/g COD/d. *Abbreviations: organic loading rate (OLR), chemical oxygen demand (COD), volatile fatty acid (VFA), root mean squared error (RMSE). 3.5.2. Influent COD and OLR were determined to be key influences on reactor biodiversity and performance The average impact of features on model output was measured using SHAP values. The top five most important features affecting COD removal included influent COD, time, OLR, temperature, and pH ( Figure 8b ). Influent COD was the most significant feature, with the highest SHAP value ( Figure 8b ). The most influential factors affecting biogas production were time, temperature, pH , Oscillibacter and influent COD ( Figure 8c ). The horizontal placement of these lines reflects the SHAP values, offering a numerical assessment of their impact on the model’s results. Analysing the distribution along the x-axis reveals that higher temperature values are associated with lower cumulative biogas and decreased reactor performance ( Figure 8c ). Two Clostridium species, temperature, time, pH were the top five factors influencing specific daily biogas ( Figure 8d ). Overall, the predictive models emphasised the importance of operational parameters on reactor performance, and biological genera within Firmicutes ( Clostridium and Oscillibacter ) on biogas production ( Figure 8b-d ). Dissimilarity analysis revealed that influent COD and OLR were the most influential operational parameters on beta diversity ( Figure 8a ). Influent butyric acid content, and mannitol, arabitol, and maltitol concentrations strongly influenced beta diversity ( Figure 8a ). Group 3 reactors were distributed across three main clusters, which appeared to be strongly influenced by individual sugars and sugar alcohols. Group 1 formed one central cluster and was strongly influenced by total sugars and sugar alcohol content ( Figure 8a ). Influent isobutyric, butyric, and acetic acid content appeared to strongly influence diversity ( Figure 8a ). Notably, individual VFAs appeared to have a more substantial influence on beta diversity than total VFA content, highlighting the importance of detailed chemical analysis for understanding microbiome diversity ( Figure 8a ). Feedstock composition was identified as a critical driver of biological parameters across all reactors. The use of ML models, particularly random forests, has proven highly effective in this work. The agreement between the dissimilarity analysis and ML models emphasised the robustness of the ML models and highlighted that influent COD and OLR were highly influential on microbiome diversity and reactor performance ( Figure 8 ). 3.6. Discussion Biological repeats of various anaerobic digestion reactors fed with MFWW were monitored over a time series to evaluate the effect of feedstock composition on the microbial community underpinning anaerobic digestion across different operational parameters and time scales. Influent COD and OLR were also identified as key influences on reactor performance by ML and dissimilarity analysis. Dissimilarity analysis showed strong clustering of reactor type by operational parameters, implying a strong impact on microbial diversity ( Figure 8a ). Additionally, individual VFAs were found to have a stronger influence on beta diversity than pH or total VFA, as did individual sugars and sugar alcohols compared to total sugar and sugar alcohol concentrations ( Figure 8a ). Isobutyric acid content in particular was found to have a major influence on microbiome biodiversity, as were influent acetic, isobutyric, and butyric acid levels ( Figure 8a ). Bioreactor chemistry highlighted distinct VFA profiles between runs 1 and 2 ( Figure 4c ). VFA profiles in run 1 consisted of a majority of isobutyric acid, which was correlated to the relative abundance of Methanomassiliicoccus 32 . The detection of both acetoclastic and hydrogenotrophic methanogens in various reactors underscores the flexibility and resilience of AD microbial communities 34 . It is important to note that Methanomassiliicoccus lack the Wood-Ljungdahl pathway for methyl group oxidation to carbon dioxide 35 . Methanomassiliicoccus have instead been found to use hydrogen-dependent methylotrophic methanogenesis pathways, utilising methylamines and methanol, which highlights an unusual metabolic pathway compared to classic hydrogenotrophic or acetoclastic methanogens 36 , 38 . Additionally, Methanomassiliicoccus have been associated with reduced biogas production 33 , and were significantly correlated with syntrophic bacteria only in group 5, which had the lowest biogas production ( Figure 6d ). Therefore, this study provides further evidence to support the hypothesis that Methanomassiliicoccus is associated with butyric acid and reduced biogas production. Moreover, this study has also highlighted the importance of Firmicutes in relation to reactor performance. The relative abundance of Firmicutes was elevated and showed little conservation over time in run 1 compared to run 2 samples. This result implies that a relatively low abundance of Firmicutes and a dynamic Firmicutes population was associated with increased biogas production. Furthermore, ML models identified three genera of Firmicutes ( Oscillibacter , and two Clostridium ) as primary factors that influence biogas production ( Figure 8c-d ). These combined results implicate Firmicutes as key bacteria for influencing in biogas production ( Figure 5 , Figure 8 ). A shift in taxonomic profile observed between days 43 and 50 for almost all reactors corresponded to increased influent COD and increased OLR ( Figure 4c , Figure 5 ). Dissimilarity analysis and ML outcomes revealed that OLR had the most substantial effect on microbiome diversity, and that influent COD was a key factor in modifying reactor performance ( Figure 8 ). The shift in taxonomy was associated with improved performance and microbiome diversity stability, indicating that flexibility within the microbiome improves performance in response to fluctuating feedstock composition ( Figure 4b , Figure 5 ). Despite having identical feedstock (MFWW), operational conditions and starting inoculum, CSTR3.1 had a distinct taxonomic profile compared to other single-stage reactors but comparable biogas production and COD removal ( Figure 4b , Figure 5 ). This result indicates that reactor COD removal and biogas production was decoupled from the microbiome profile. Syntrophic relationships showed stronger correlations in single- and first-stage compared to second-stage reactors ( Figure 6d ). Additionally, network analysis revealed that syntrophic species had higher node connectivity in mesophilic compared to thermophilic reactors ( Figure 6 ). Several bacterial genera are involved in syntrophic relationships essential for AD. Syntrophobacter converts butyrate and propionate to acetate, which is then utilised by methanogens such as Methanosarcina and Methanosaeta 37 . Similarly, Syntrophomonas wolfei degrades butyrate in association with hydrogen-utilising methanogens, maintaining low hydrogen partial pressure, which is critical for the energetics of the reaction 38 . Additionally, Smithella species grow on propionate only in the presence of methanogenic bacteria that remove hydrogen and formate, facilitating propionate oxidation 36 . Candidatus Cloacamonas , part of the WWE1 candidate phylum, and Syntrophomonas are involved in the fermentation of amino acids and butyrate, respectively, indicating their significant roles in syntrophic degradation processes ( Figure 1 ) 39 . Furthermore, the presence of syntrophic bacteria like Thermotoga and Thermacetogenium in thermophilic digesters supports syntrophic acetate oxidation under high-temperature and high-ammonia conditions, highlighting the adaptability and importance of these bacteria in diverse AD environments 27 . Additionally, Cytophaga , Herbaspirillum , Symbiobacterium , Comamonas , and Allochromatium have been associated with enhanced biogas production, although their exact roles remain unclear 40 , 41 . A greater understanding of the microbiome underpinning AD could improve reactor stability and reduce start-up times 28 , 42 , 43 . The well-described process of acetoclastic methanogenesis has formed the basis for much top-down AD microbiome optimisation. However, our results agree with the findings that the obligate acetoclastic methanogen Methanosaeta has been associated with reactor dysfunction, and acetoclastic methanogenesis is not always the dominant pathway for methane production 41 , 43 – 45 . 4. Conclusions and Future Directions This study investigated the effect of complex feedstock (MFWW) from the microbial protein industry on microbiome diversity and reactor performance in a time series. ML models and dissimilarity analysis highlighted the importance of crude and detailed chemical analysis for influencing reactor biodiversity and performance. The findings also indicated that Methanomassiliicoccu s is associated with high butyric acid concentrations and reduced biogas production, providing new insights into the newly discovered methanogen. Finally, Firmicutes were implicated in reactor performance and were found to be highly dynamic in reactors with high biogas production. Future research should focus on building upon the findings of this study by elongating the run times, including untargeted chemical analysis and additional -omics tools. HPLC analysis was limited to known compounds, including VFAs and sugars and sugar alcohols, which failed to identify all compounds of interest within the reactors. Untargeted analysis would be recommended to identify all compounds of interest within the reactors. Furthermore, individual compound manipulation of characterised feedstocks (MFWW) could be conducted to validate the findings of the dissimilarity analysis and ML models. Preliminary exploration of bioreactor chemistry and operational parameters as predictors of microbiome structure was used to investigate whether microbial community assembly is deterministic or stochastic (Supplementary Materials 3.3). However, more data will be required to further explore this topic. Moreover, advanced time series analysis was challenging to implement. An extended run time of at least 6 months would be recommended to further explore the temporal trends and whether community can be predicted based on feedstock composition and operational conditions. Finally, we recommend additional -omics to investigate the role of structure and function in relation to reactor chemistry and performance parameters. Meta-transcriptomics is a common approach used in microbiome studies to elucidate the functional profile of the microbiome and should be applied to further explore some interesting relationships observed during the experimental run, such as the metabolic basis for the elevated production of valeric acid in CSTR3.1. A meta-pan genomic approach could also be applied to compare reactor metagenomes and their relationship to bioreactor chemistry. This work developed robust predictive ML models based on AD chemical and biological fingerprinting. The combined ML approach derived actionable insights for the potential of Firmicutes , Methanomassiliicoccus and butyric acid as biological and chemical markers to monitor and predict system performance. By integrating biological, chemical, and computational methods, researchers can develop robust monitoring methods that consider whole systems approaches to enhance AD performance. Detailed feedstock analysis in tandem with the application of metagenomic sequencing and reactor chemical analysis are key methods for capturing the complexity of these communities. By investigating substrates ranging from single molecules to complex macroparticles, we may be able uncover microbial structure and function trends. By effectively managing operational parameters and bioreactor chemistry, it is possible to harness the full potential of AD, ensuring efficient wastewater treatment and resource recovery. Data Availability All sequencing data have been submitted to the European Nucleotide Archive (ENA) under the project ID PRJEB80086. Sequences reported in this study are deposited in the EMBL-EBI database under accession numbers ERS22193437-ERS22335023. The correspondence between accession numbers, sequences and metadata is provided in Supplementary Database 1. Machine learning models are open source and available at https://github.com/MGuo-Lab/AnaerobicDigestionML . Declaration of competing interests The authors declare no known conflicts of interest. Author contributions EP and MG conceived the study. EP acquired the data. EP and XS conducted the experiments, analysed and interpreted the data. EP, XS, PE, MT and MG drafted or revised the manuscript. Acknowledgements We are grateful to Marlow Ingredients and the research and development team, especially Dr Mark Taylor, for supplying the feedstock for the experiments. EP and MG would like to acknowledge the UK Engineering and Physical Sciences Research Council (EPSRC) and Monde Nissin Corporate for providing fundings under the EPSRC iCASE programme. Footnotes https://github.com/MGuo-Lab/AnaerobicDigestionML References 1. ↵ Tariq , A. & Mushtaq , A . Untreated wastewater reasons and causes: a review of most affected areas and cities . Int. J. Chem. Biochem. Sci 23 , 121 – 143 ( 2023 ). OpenUrl 2. ↵ Jones , E. R. , Van Vliet , M. T. , Qadir , M. & Bierkens , M. F . Country-level and gridded estimates of wastewater production, collection, treatment and reuse . Earth System Science Data 13 , 237 – 254 ( 2021 ). OpenUrl CrossRef 3. ↵ Dutta , D. , Arya , S. & Kumar , S . Industrial wastewater treatment: Current trends, bottlenecks, and best practices . Chemosphere 285 , 131245 ( 2021 ). OpenUrl CrossRef PubMed 4. ↵ Soares , A. Wastewater treatment in 2050: Challenges ahead and future vision in a European context . Environmental Science and Ecotechnology 2 , 100030 ( 2020 ). OpenUrl CrossRef 5. ↵ De Vrieze , J. & Verstraete , W . Perspectives for microbial community composition in anaerobic digestion: from abundance and activity to connectivity . Environmental microbiology 18 , 2797 – 2809 ( 2016 ). OpenUrl CrossRef 6. ↵ Shoener , B. , Bradley , I. , Cusick , R. & Guest , J . Energy positive domestic wastewater treatment: the roles of anaerobic and phototrophic technologies . Environmental Science: Processes & Impacts 16 , 1204 – 1222 ( 2014 ). OpenUrl CrossRef PubMed 7. ↵ Wu , D. et al. Anaerobic digestion: A review on process monitoring . Renewable and Sustainable Energy Reviews 103 , 1 – 12 ( 2019 ). OpenUrl CrossRef 8. ↵ Madsen , M. , Holm-Nielsen , J. B. & Esbensen , K. H . Monitoring of anaerobic digestion processes: A review perspective . Renewable and sustainable energy reviews 15 , 3141 – 3155 ( 2011 ). OpenUrl CrossRef 9. ↵ Mercado , J. V. , Koyama , M. & Nakasaki , K . Short-term changes in the anaerobic digestion microbiome and biochemical pathways with changes in organic load . Science of The Total Environment 813 , 152585 ( 2022 ). OpenUrl CrossRef PubMed 10. ↵ Rutland , H. , You , J. , Liu , H. , Bull , L. & Reynolds , D . A systematic review of machine-learning solutions in anaerobic digestion . Bioengineering 10 , 1410 ( 2023 ). OpenUrl CrossRef PubMed 11. ↵ Long , F. , Wang , L. , Cai , W. , Lesnik , K. & Liu , H . Predicting the performance of anaerobic digestion using machine learning algorithms and genomic data . Water Research 199 , 117182 ( 2021 ). 12. ↵ Wang , L. , Long , F. , Liao , W. & Liu , H . Prediction of anaerobic digestion performance and identification of critical operational parameters using machine learning algorithms . Bioresource technology 298 , 122495 ( 2020 ). 13. ↵ Geypens , M. , Vanderdriech , E. , Goncalves , L. , Kcfs , M. & Sasaki , c. Apha , AEG; Awwa , ADE ; WEF, LSC Standard Methods for the Examination of Water and Wastewater . Washington DC ; American Public Health Association , 1995. Atlas , RM ; Bartha , R. Microbial Ecology–Fundamentals and Applications . New York : Addison Wesey Longman In. FRANCIHELE CARDOSO MÜLLER 42 ( 2012 ). 14. ↵ Schloss , P. D. et al. Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities . Applied and environmental microbiology 75 , 7537 – 7541 ( 2009 ). OpenUrl Abstract / FREE Full Text 15. ↵ Quast , C. et al. The SILVA ribosomal RNA gene database project: improved data processing and web-based tools . Nucleic acids research 41 , D590 – D596 ( 2012 ). OpenUrl CrossRef PubMed Web of Science 16. ↵ Letunic , I. & Bork , P . Interactive Tree Of Life (iTOL) v5: an online tool for phylogenetic tree display and annotation . Nucleic acids research 49 , W293 – W296 ( 2021 ). OpenUrl CrossRef PubMed 17. ↵ Deng , Y. et al. Molecular ecological network analyses . BMC bioinformatics 13 , 1 – 20 ( 2012 ). OpenUrl CrossRef PubMed 18. ↵ Shannon , P. et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks . Genome research 13 , 2498 – 2504 ( 2003 ). OpenUrl Abstract / FREE Full Text 19. ↵ Feranchuk , S. et al. Tools and a web server for data analysis and presentation in microbial ecology . Community Ecology 20 , 230 – 237 ( 2019 ). OpenUrl CrossRef 20. ↵ Veolia. Wastewater characterisation summary . ( 2013 ). 21. ↵ Quorn . Research Brief confidential: assessment of upfront technology in the capture of proteins within quorn fermenter waste streams . ( Marlow foods , 2016 ). 22. ↵ Environment Agency . (ed Environment Agency ) ( Gov.uk , https://www.gov.uk/government/publications/waste-water-treatment-works-treatment-monitoring-and-compliance-limits/waste-water-treatment-works-treatment-monitoring-and-compliance-limits , 2019 ). 23. ↵ Conrado , L. et al. Anaerobic Conversion of Proteinogenic Amino Acids When Methanogenesis Is Inhibited: Carboxylic Acid Production from Single Amino Acids . Fermentation 10 , 237 ( 2024 ). 24. Wainaina , S. , Lukitawesa , Kumar Awasthi , M. & Taherzadeh , M. J. Bioengineering of anaerobic digestion for volatile fatty acids, hydrogen or methane production: a critical review . Bioengineered 10 , 437 – 458 ( 2019 ). OpenUrl CrossRef PubMed 25. ↵ Yao , H. , Flanagan , B. M. , Williams , B. A. , Mikkelsen , D. & Gidley , M. J . Factors affecting plant food particle behaviour during in vitro colonic microbial fermentation . Trends in Food Science & Technology , 104802 ( 2024 ). 26. ↵ Liu , H. et al. Effects of different amino acids and their configurations on methane yield and biotransformation of intermediate metabolites during anaerobic digestion . Journal of environmental management 296 , 113152 ( 2021 ). 27. ↵ Zamanzadeh , M. , Hagen , L. H. , Svensson , K. , Linjordet , R. & Horn , S. J . Anaerobic digestion of food waste–effect of recirculation and temperature on performance and microbiology . Water Research 96 , 246 – 254 ( 2016 ). OpenUrl CrossRef 28. ↵ Louca , S. et al. Function and functional redundancy in microbial systems . Nature ecology & evolution 2 , 936 – 943 ( 2018 ). doi: 10.1038/s41559-018-0519-1 OpenUrl CrossRef PubMed 29. ↵ Dalantai , T. et al. Complex network analysis of slaughterhouse waste anaerobic digestion: From failure to success of long-term operation . Bioresource Technology 361 , 127673 ( 2022 ). 30. ↵ Xu , R. et al. Organic loading rate and hydraulic retention time shape distinct ecological networks of anaerobic digestion related microbiome . Bioresource Technology 262 , 184 – 193 ( 2018 ). OpenUrl CrossRef PubMed 31. ↵ Zhou , J. et al. Different organic loading rates on the biogas production during the anaerobic digestion of rice straw: A pilot study . Bioresource technology 244 , 865 – 871 ( 2017 ). OpenUrl CrossRef PubMed 32. ↵ Nikitina , A. A. et al. Syntrophic butyrate-oxidizing consortium mitigates acetate inhibition through a shift from acetoclastic to hydrogenotrophic methanogenesis and alleviates VFA stress in thermophilic anaerobic digestion . Applied Sciences 13 , 173 ( 2022 ). 33. ↵ Lianhua , L. et al. Anaerobic co-digestion of Pennisetum hybrid and pig manure: A comparative study of performance and microbial community at different mixture ratio and organic loading rate . Chemosphere 247 , 125871 ( 2020 ). 34. ↵ Ziganshin , A. M. , Liebetrau , J. , Pröter , J. & Kleinsteuber , S . Microbial community structure and dynamics during anaerobic digestion of various agricultural waste materials . Applied Microbiology and Biotechnology 97 , 5161 – 5174 ( 2013 ). doi: 10.1007/s00253-013-4867-0 OpenUrl CrossRef 35. ↵ Borrel , G. , Adam , P. S. & Gribaldo , S . Methanogenesis and the Wood–Ljungdahl pathway: an ancient, versatile, and fragile association . Genome biology and evolution 8 , 1706 – 1711 ( 2016 ). OpenUrl CrossRef PubMed 36. ↵ Tong , J. et al. Microbial community evolution and fate of antibiotic resistance genes during sludge treatment in two full-scale anaerobic digestion plants with thermal hydrolysis pretreatment . Bioresource Technology 288 , 121575 ( 2019 ). doi: 10.1016/j.biortech.2019.121575 OpenUrl CrossRef 37. ↵ Cai , M. et al. Metagenomic reconstruction of key anaerobic digestion pathways in municipal sludge and industrial wastewater biogas-producing systems . Frontiers in Microbiology 7 , 778 ( 2016 ). 38. ↵ Stams , A. J. & Plugge , C. M . Electron transfer in syntrophic communities of anaerobic bacteria and archaea . Nature Reviews Microbiology 7 , 568 – 577 ( 2009 ). OpenUrl CrossRef PubMed Web of Science 39. ↵ Sträuber , H. , Lucas , R. & Kleinsteuber , S . Metabolic and microbial community dynamics during the anaerobic digestion of maize silage in a two-phase process . Applied Microbiology and Biotechnology 100 , 479 – 491 ( 2016 ). doi: 10.1007/s00253-015-6996-0 OpenUrl CrossRef 40. ↵ Gonzalez-Martinez , A. , Garcia-Ruiz , M. J. , Rodriguez-Sanchez , A. , Osorio , F. & Gonzalez-Lopez , J . Archaeal and bacterial community dynamics and bioprocess performance of a bench-scale two-stage anaerobic digester . Applied microbiology and biotechnology 100 , 6013 – 6033 ( 2016 ). OpenUrl CrossRef 41. ↵ Castellano-Hinojosa , A. , Armato , C. , Pozo , C. , González-Martínez , A. & González-López , J . New concepts in anaerobic digestion processes: recent advances and biological aspects . Applied microbiology and biotechnology 102 , 5065 – 5076 ( 2018 ). doi: 10.1007/s00253-018-9039-9 OpenUrl CrossRef 42. ↵ Demirel , B. & Scherer , P . The roles of acetotrophic and hydrogenotrophic methanogens during anaerobic conversion of biomass to methane: a review . Reviews in Environmental Science and Bio/Technology 7 , 173 – 190 ( 2008 ). OpenUrl CrossRef 43. ↵ Lawson , C. E. et al. Common principles and best practices for engineering microbiomes . Nature Reviews Microbiology 17 , 725 – 741 ( 2019 ). doi: 10.1038/s41579-019-0255-9 OpenUrl CrossRef 44. Mulat , D. G. et al. Quantifying contribution of synthrophic acetate oxidation to methane production in thermophilic anaerobic reactors by membrane inlet mass spectrometry . Environmental science & technology 48 , 2505 – 2511 ( 2014 ). OpenUrl PubMed 45. ↵ Mosbæk , F. et al. Identification of syntrophic acetate-oxidizing bacteria in anaerobic digesters by combined protein-based stable isotope probing and metagenomics . The ISME journal 10 , 2405 – 2418 ( 2016 ). OpenUrl CrossRef PubMed View the discussion thread. Back to top Previous Next Posted January 13, 2025. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Temporal Dynamics of Microbial Communities in Anaerobic Digestion: Influence of Temperature and Feedstock Composition on Reactor Performance and Stability Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Temporal Dynamics of Microbial Communities in Anaerobic Digestion: Influence of Temperature and Feedstock Composition on Reactor Performance and Stability Ellen Piercy , Xinyang Sun , Peter R. Ellis , Mark Taylor , Miao Guo bioRxiv 2025.01.12.632589; doi: https://doi.org/10.1101/2025.01.12.632589 Share This Article: Copy Citation Tools Temporal Dynamics of Microbial Communities in Anaerobic Digestion: Influence of Temperature and Feedstock Composition on Reactor Performance and Stability Ellen Piercy , Xinyang Sun , Peter R. Ellis , Mark Taylor , Miao Guo bioRxiv 2025.01.12.632589; doi: https://doi.org/10.1101/2025.01.12.632589 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Microbiology Subject Areas All Articles Animal Behavior and Cognition (7622) Biochemistry (17650) Bioengineering (13871) Bioinformatics (41881) Biophysics (21424) Cancer Biology (18566) Cell Biology (25461) Clinical Trials (138) Developmental Biology (13365) Ecology (19866) Epidemiology (2067) Evolutionary Biology (24290) Genetics (15590) Genomics (22476) Immunology (17713) Microbiology (40331) Molecular Biology (17148) Neuroscience (88473) Paleontology (666) Pathology (2827) Pharmacology and Toxicology (4816) Physiology (7635) Plant Biology (15114) Scientific Communication and Education (2044) Synthetic Biology (4286) Systems Biology (9815) Zoology (2268)