Kefir4All, a citizen science initiative to raise awareness of the roles that microbes play in food fermentation

A total of 102 citizen scientists were recruited from across 9 school settings and were referred to as school groups, while 21 citizen scientists were recruited outside of school settings and were referred to as “non-school group—milk kefir” or “non-school group—water kefir” based on the kefir type they were fermenting during the project. At the start of the project, the 123 citizen scientists received a milk kefir or water kefir grain (each originating from a common source), the required fermentation equipment (details below), and fermentation guidelines. Milk or water kefir was then produced by the citizen scientists in private households for up to 21 weeks. Production of the milk or water kefir required the “transfer” of the kefir grain to fresh medium every 1–3 days. The milk kefir medium represented different types of milk, specifically citizen scientists could choose to ferment in a) full fat milk, b) low fat milk, c) goat milk or d) skimmed/semi skimmed milk, while the water kefir medium was comprised of different combinations of sugary water, with or without additional fruits, specifically citizen scientists could choose to ferment in a) 12 gram of white sugar and 1 dried fig, b) 12 gram of white sugar, 1 dried fig and 1 slice of fresh lemon, c) 12 gram of white sugar and 1 dried apricot or d) 12 gram of brown sugar. Kefir grains and liquids were collected at week 1, week 5, week 9, week 13, week 17, and week 21, and metadata (such as information about storage conditions, type of milk or sugar water used, and frequency with which the grains were “transferred”) were reported after each fermentation by citizen scientists using the Survey Monkey platform (16).

Citizen scientists were recruited during a series of fermentation workshops, given in-person in a school for the school group or online for the non-school group. The workshop was divided into (i) a 40-minute talk (Introduction to microbiology, Introduction to DNA sequencing, and Introduction to fermented foods), (ii) a 20-minute workshop on kefir fermentation, followed by an opportunity for questions from the potential citizen scientists, and (iii) distribution of the kefir fermentation kits. The workshop was used to provide background scientific knowledge about microbiology, fermentation, and bioinformatics, which aided in the explanation of the research project and the role of the citizen scientists, with an emphasis on how their contributions would aid the research project and its objectives. Citizen scientists further received recordings of the fermented food workshop and steps involved in sample collection to use as supporting materials throughout the project.

Citizen scientists received fermentation kits that included a kefir grain, fermentation vessels, thermometers, utensils, pH strips, collection tubes, and fermentation guidelines. Fermentation equipment was purchased from Kefirako (Borgla d.o.o.,Valvasorjeva ulica 40, 2000 Maribor, Slovenija), including the fermentation jars, pH test strips, and thermometers. Collection tubes were purchased from Thermo Scientific, UK. The milk kefir grains were sourced from Every Good Thing, UK (https://everygoodthing.co.uk/products/milk-kefir-grains). The water kefir grains were sourced from Live Ferments, Ireland (https://kefirgrains.ie/). Fermentation guidelines generated by researchers conducting the study included detailed written instructions on the steps involved in the fermentation process and in sample collection, recordings of the workshop (https://youtube.com/playlist?list=PLlgXUwWwJdfm5KcWpFpa0f2pO0NxohG3T), and links to surveys used to collect metadata (see Supplementary Materials).

The Kefir4All project and the associated engagement activities were evaluated through a series of surveys. Our study included four surveys: “Workshop Survey,” “Fermentation Survey,” “Project Completion Survey—Citizen Scientists,” and “Project Completion Survey—Teachers” (see Supplementary Materials), all constructed using Survey Monkey (16). The citizen scientists were asked to complete the “Workshop Survey,” at the end of our recruitment workshop, which included opinions provided by members of the public who did not subsequently take part in the Kefir4All study. The Workshop Survey examined the self-reported educational impact on attendees as a result of attending a 1-hour workshop. The Workshop Survey also assessed the pre-existing microbiology and fermentation-related knowledge of the participants. The Fermentation Survey was used to collect metadata about each fermentation process completed by the citizen scientists. The Project Completion Survey—Citizen Scientists and the Project Completion Survey—Teachers primarily examined the self-reported educational impact on the citizen scientists, as a result of taking part in the project, and the personal accounts of the citizen scientists and teachers, respectively, relating to strengths of the project and/or suggested improvements. Citizen scientists were asked to complete the Fermentation Survey throughout the project and the Project Completion Survey—Citizen Scientists at the end of the project. School-based coordinators (teachers) of the Kefir4All project were asked to complete the Project Completion Survey—Teachers at the end of the project. All survey responses were analyzed qualitatively.

To facilitate continuous participation with, and maintain interest from the citizen scientists, a number of activities and challenges were devised. Six career events (three speakers per event) were arranged, which the citizen scientists could attend virtually and hear the accounts and experiences of researchers at different stages of their careers. To encourage continued participation of the citizen scientists in the project, a leader board was circulated every week, highlighting the citizen scientist group with the most activity. Standardization by number of citizen scientists was not considered, given the larger number of “non-fermenters” in larger school groups. The leader board also identified the “citizen scientist of the week,” the one who completed the most fermentations that week (based on number of surveys they completed). The citizen scientists received goodie bags containing beanie hats, cups, and stationary upon providing their final set of samples. Citizen scientists were able to continuously engage with and avail of the expertise of the researchers throughout the project via email or in person during sample collection and return school visits. The school-based citizen scientists were invited to the Teagasc Food Research Centre, Moorepark for a “Kefir Day” in May 2022. During this visit, the citizen scientists took part in tours of the research facilities, short laboratory-based workshops, and visited scientific stands covering a number of topics such as bacteriophage, food fermentation, the gut microbiome, and next-generation sequencing. Completion certificates and preliminary data reports, including the microbial composition of the participant’s kefir grains and liquid samples, were made available to each contributing citizen scientist at the end of the project. Preliminary data were presented in a PDF document, together with a recap of the conceptions discussed during the project.

Milk kefir and water kefir shotgun metagenomic sequences generated in Kefir4All or acquired from Walsh et al. (17) and Breselge et al. (unpublished data) were processed using the Teagasc high-performance computing cluster. Sequences from Walsh et al. (17) and Breselge et al. (unpublished data) were included to compare compositional profiles between metagenomes acquired using a more traditional laboratory-based approach and a citizen science approach. The laboratory-based approach included the fermentation and sampling of milk and water kefir under standardized conditions in a laboratory setting. Initially, raw paired-end FASTQ files containing the metagenomic shotgun sequences were trimmed using Trimgalore (v.0.6.1) (18) to remove adapter sequences and low-quality reads (average quality score <Q₂₀), fragmented (<75 bp) and with more than two ambiguous nucleotides. Host-contaminating reads were removed, through an initial alignment of metagenomes against a reference database, followed by read removal of any aligned reads. Specifically, milk kefir metagenomes were searched against a host reference database containing a bovine, phiX 174, and human genome using Bowtie2 (v.2.3.4) (19) with the parameter “sensitive-local.” Water kefir metagenomes were searched against a reference database containing a phiX 174 and human genome (GRCh38) using Bowtie2 (v2.3.4) with the parameter “sensitive-local.” Reference genomes were downloaded from iGenomes (https://support.illumina.com/sequencing/sequencing_software/igenome.html) and represented the most up to date versions available in NCBI (November 2022). Remaining high-quality reads were sorted and split to create forward, reverse, and unpaired reads output FASTQ formatted files for each metagenome. Compositional analysis was performed with MetaCache (v 2.3) (20).

Analysis of survey data from each of the outlined survey was completed in R-4.0.2 (21). Correlation analysis was calculated using the Spearman correlation method (22) through the ggscatter function included in ggpubr (v0.4.0). Frequency counts for all survey responses collected in this study were described in a contingency table using the stats package (3). The Fisher’s exact test was used to determine significant associations between frequency counts of categorical survey responses in the Workshop Survey. The chi-square test of independence was used to determine significant associations between frequency counts of the categorical variables “Fermentation characteristics” and “Fermentation categories,” “Grain wash,” “Room temperature,” “pH,” “Grain weight,” “Water type,” “Liquid volume,” and “Sugar weight,” all of which were collected in the Fermentation Survey. P-values obtained using the Fisher’s exact test and the chi-square test were adjusted using the Benjamani-Hochberg correction to account for multiple comparisons. Beta diversity values were calculated using the vegdist function of the vegan (v2.6.2) package (23). The fviz_nbclust function included in factoextra (v1.0.7) (24) was used to predict the optimal number of clusters using the average silhouette method. Cluster validation was performed using the Dunn index test (25) through the Dunn function in clValid (v 0.7). Hierarchical clustering was performed using the hclust function using the ward.D2 agglomeration method in stats (v4.1.2) (21). Bootstrapping was applied to assess cluster stability using the clusterboot function of the fpc (v2.2.10) package (26). Jaccard coefficient values obtained using bootstrapping were used to evaluate cluster stability; values <0.6 were considered unstable, and values >0.85 were considered highly stable. Bar charts were produced from survey data using the geom_segment function of the ggplot2 (v3.3.6) package (27). Word clouds were produced using the geom_text_wordcloud_area function in the ggwordcloud package (v0.5.0) (28). The ordination plot was produced from Bray-Curtis dissimilarity beta diversity values between milk and water kefir metagenomes using the geom_point function of the ggplot2 (v3.3.6) package.

The Kefir4All project involved a series of steps (Fig. 1) that began with the organization and advertisement of workshops (see Materials and Methods), including letters being sent to local secondary schools to encourage their participation and posts on social media. The citizen scientists were recruited at the end of a 1-hour workshop (see Materials and Methods), primarily given in school settings. At the beginning of this workshop, the potential citizen scientists were asked to complete questions 2–15 of the Workshop Survey to assess their self-reported background knowledge and interest relating to the topics of microbiology and fermentation. The remaining questions (questions 1 and 16–19) of the Workshop Survey were completed at the end of the workshop to assess if attendance at the workshop raised awareness and interest in potential participants of the topics microbiology and fermentation (Fig. 1; Fig. S1).

The Workshop Survey was completed by 87 individuals, 58 (67%) of whom volunteered for the Kefir4All project (Fig. S1A; Fig. 2A and B). The survey responses collected from responders who took part in the project generally followed the same patterns as the survey responses from responders who did not take part in the project, with two notable exceptions, i.e., participants were more likely to take part in the project if they had a greater interest in fermented foods (Fig. 2A) and/or microbiology (Fig. 2B) (Fisher’s exact test, adjusted P-values <0.01).

Of the total responders, 25 (29%) had an interest in microbiology, few were disinterested in microbiology (n = 5, 5%), and approximately half of the responders (n = 43, 50%) had a neutral interest (Fig. 2B). A neutral interest was the most selected option in responders who agreed to take part in the project and those who did not. However, more of the non-participating responders selected the option “I don't know what it means” (30% of the total non-participating responders) compared to participating responders (7% of the total participating responders) (Fig. 2B). Most responders felt they lacked background knowledge of microbiology, approximately half of the responders (n = 41, 47%) considered that they had a little background knowledge, and a further 28 (32%) responders felt they had no background knowledge, while only one (1%) and 11 of the 87 responders (13%) regarded themselves as knowledgeable and very knowledgeable, respectively (Fig. S1C). Also, 64 responders (74%) expressed an interest in fermented foods (Fig. S1D) despite many responders “never” consuming fermented foods (Fig. S1F) and the majority having no previous experience of fermentation (n = 72, 84%) (Fig. S1G). Knowledge gaps were also evidenced by the fact that, prior to the workshop, 37 responders (45%) were not aware that foods they previously consumed were fermented. Additionally, a small minority of responders (n = 8, 14%) referred to unfermented foods such as milk (n = 4) or alternatively stated that they were not sure (n = 3) when asked the question, what fermented foods do you eat?

While many of the non-participating responders expressed an interest in fermented food (45%), the remaining non-participating responders selected “Not so interested” (37%) or “Not at all interested” (19%), which was in stark contrast to the participating responders with few selecting the “Not so interested” (9%) or “Not at all interested” (2%) options (Fig. 2A). We examined the association between interest in fermented foods and consumption of fermented foods. When asked how often they consumed fermented foods, 32 (37%) of the responders selected “Never,” 21 (25%) selected “Occasionally,” 22 (26%) selected “Regular,” and 9 (9%) selected “Daily.” Of those who declared that they had previously consumed fermented food, only 33% of responders (n = 18) enjoyed the sour taste of the food. The responders who “Never” consumed fermented food (Fig. S1F) also tended to select “Somewhat interested” (n = 13, 34%), “Not so interested” (n = 10, 67%), or “Not at all interested” (n = 5, 83%) in fermented foods. In contrast, most “Regular” and “Daily” consumers of fermented food expressed an interest in fermentation with only one (17%) daily consumer and two (13%) regular consumers expressing a disinterest in fermented foods. Similarly, only one (11%) daily consumer and seven (32%) regular consumers were unaware that some of the foods they had consumed were fermented foods prior to the workshop compared to 18 (62%) of those who reported that they had not previously consumed fermented foods. However, known prior consumption of fermented food did not associate with taking part in the project (Fisher’s exact test, P-value >0.05), with 19 of the total 32 responders (52%) agreeing to take part in the project despite never, to their knowledge, consuming fermented foods.

We then examined the association between interest in fermented foods/microbiology and subject choices among students in second-level education of relevance to the citizen science project, specifically if the responder chose to study the topics of biology (n = 36, 41%), home economics (n = 8, 9%), both (n = 36, 30%), or neither subject (n = 16, 19%). It was established that subject choice was not associated with interest in fermented foods (Fisher’s exact test, P-value >0.05), with a higher portion of responders (n = 5, 31%) who did not take either subject expressing an interest in fermented foods. Subject choice did associate with interest in microbiology (Fisher’s exact test, P-value >0.05); however, many of the responders who selected the option “I don’t know what it means” when asked about their interest in microbiology had selected biology (n = 10, 83%) as part of their studies. Furthermore, subject choice did not associate with participation rates in the project (Fisher’s exact test, P-value >0.05), with similar subject choice profiles between those who participated in the project and those who did not. Indeed, 13 responders (81%) who did not select biology and/or home economics as part of their studies took part in the study, accounting for 22% (n = 13) of the recruited responders. When considering where the responders source their scientific information, the majority of the responders used YouTube (n = 34, 39%), followed closely by information received through schools (n = 30, 34%), with few responders using websites (n = 9, 10%) and/or books (n = 3, 3%) (Fig. S1I). However, responders who utilized books and websites tended to engage with the information provided with only one responder (11%) who selected “websites” choosing to never engage with those sources of information. Six responders (20%) engaged with scientific information in schools every day, and a further 12 (40%) responders engaged on a regular basis with scientific information provided in schools. Those who identified schools as the primary source of their scientific knowledge also had higher levels of daily and regular engagement compared to YouTube, which while having lower proportions of no engagement had higher portions of occasional engagement compared to schools.

Also, 71 of the 87 (82%) responders found the workshop to be interesting/very interesting, with 77 responders (89%) reporting an improved interest and general knowledge of microbiology and/or fermented foods as a result of the workshop. Specifically, 54 responders (62%) reported an improved knowledge relating to both microbiology and fermented foods, 18 reported an improvement solely in microbiology, and 5 in fermented foods; only 9 of the 87 (10%) responders felt no benefit as a result of attending the workshop (Fig. S1J). The majority of the responders who did not feel they benefited from the workshop (Fig. S1J) expressed an absence of interest in fermentation and/or an absence of background knowledge in microbiology. Of these nine responders, four (50%) were “not at all interested in fermentation,” and five (62%) selected “I don’t know what it means” when asked about background knowledge of microbiology. Despite this, of the 12 responders who did not know what microbiology was prior to the workshop, seven (58%) reported an improvement of both their interest and general knowledge of microbiology by the end of the workshop. Greater learning outcomes tended to be achieved by responders who expressed an interest in the topic microbiology (Fisher’s exact test, adjusted P-value <0.05) (Fig. 2C), which in turn had a relationship with subject choice (Fisher’s exact test, adjusted P-value <0.05) (Fig. 2D). Similarly, responders who expressed a higher interest in microbiology and/or fermentation were more likely to take part in the citizen science project (Fisher’s exact test, P-value >0.05) (Fig. 2A and B). For example, 83% of responders (n = 5) who selected “Not at all interested” when asked to rate their interest in fermented foods did not taking part in the project, while, in contrast, 81 responders (n = 17) who selected “Interested”’ and 100% of responders (n = 5) who selected “Extremely Interested” took part in the project (Fig. 2A and B). Furthermore, 75% of responders who expressed that they did not benefit from the workshop did not take part in the project, while only 23% of responders who expressed an increased interest and general knowledge of microbiology and fermented foods after attending the workshop did not take part in the project. It is important to note other parameters not considered in this study such as age and employment in the non-school group could influence learning outcomes and participation in the Kefir4all project. Such parameters capture personally identifiable information and were not considered to ensure the creation of an anonymized data set.

Kefir4All consisted of 123 participants, separated into 9 local school groups (with an average of 11 participants per school group) (n = 102) and 2 non-school groups (n = 21) separated into non-school group—milk kefir and non-school group—water kefir (Fig. 1). The participants were selected based on their interest in taking part in the project. After quality filtering (which included the removal of blank/incomplete surveys or those missing essential details such as the participants ID number), the Fermentation Survey consisted of 1,823 responses. These were provided over the course of 21 weeks (week 1–week 21) (Fig. 1 and 3B). The Fermentation Survey functioned as a fermented food log, in which citizen scientists could record information and observations about each fermentation they performed, e.g., grain growth, fermentation characteristics (bubbles, biofilm formation, etc.), and room temperature. The completion rate of these surveys provides insights into the number of citizen scientists who contributed to the project. The number of samples provided also reflected the citizen scientists’ participation, with 516 kefir samples provided by citizen scientists across 6 time points over 21 weeks (Fig. 3D). As recruitment was performed in “good faith,” with no prior knowledge of the citizen scientist’s interest in, and/or likelihood to complete the project, we initially used the Fermentation Survey to evaluate the contributions of the citizen scientists. This was judged based on the number of surveys completed by each participant. If a citizen scientist completed ≥20 Fermentation Surveys, they were considered “Regular fermenters” (n = 44). Participants were deemed “Occasional fermenters” (n = 10) if they completed greater than 10 but less than 20 surveys, “Rare fermenters” (n = 34) if they completed <10 Fermentation Surveys or “Non-fermenters” (n = 35) if no surveys were complete during the project (Fig. 3A). Across all of the 11 groups recruited, there was an average of 4 “Regular fermenters”, 1 “Occasional fermenters”, 4 “Rare fermenters”, and 3 “Non-fermenters.” We compared the fermentation categories obtained through the evaluation of the Fermentation Survey to the self-determined fermentation categories selected by citizen scientists. Self-determined fermentation categories were recorded when citizen scientists were asked “How would you rate your participation in the citizen science project?” in the evaluation survey circulated at the end of the project. Of the 15 responders who supplied a citizen scientist ID, discrepancies between both approaches were only found in 3 cases. Such discrepancies may arise as some fermentations may beperformed but not recorded in the Fermentation Survey.

We monitored the number of Fermentation Surveys completed during the Kefir4All study, separated according to the weeks of the study (Fig. 3C). During the first 5 weeks, participation was high with 214 surveys completed in the 7-day period of week 1 and 91 active participants providing 109 samples, and 502 surveys completed within the 28-day period of week 5, with 75 active participants providing 99 samples in week 5. However, the survey completion rate, the active number of participants, and the number of samples provided (Fig. 3C and D) all gradually decreased following week 5, with the lowest values recorded at week 21 with 182 surveys completed, 39 active participants, and 65 samples provided. Citizen scientists were asked to provide six kefir liquid samples and six kefir grain samples during the project (Fig. 1). On average, the citizen scientists provided four kefir liquid samples and four kefir grain samples. Correlation analysis using the Spearman method indicated that there was a positive relationship between the group size and the frequency of “Non-fermenters” (Spearman’s correlation adjusted P-value = 0.02) and “Rare fermenters” (Spearman’s correlation adjusted P-value = 0.03). Of particular interest, we noted a higher proportion of “Regular fermenters” in school groups where teachers classified their contribution as “A lot” or “A moderate amount,” compared to teachers who did not complete the evaluation survey or who selected “None at all.”

As the Kefir4All project progressed past the first 5 weeks, the pattern of decreased participation over time became apparent, and some school coordinators requested engagement activities. To address this, a leaderboard (Fig. 4) was developed and circulated weekly to the citizen scientists from second-level school and non-school settings. The leaderboard graphically displayed the collective survey responses per group that week (Fig. 4A), the collective number of surveys answered each day that week, separated according to group (Fig. 4B), and the collective number of surveys answered since the beginning of the project (Fig. 4C).

In addition, career events were organized, during which career opportunities available in multiple sectors of science, such as chemistry, microbiology, and agricultural science, were discussed, and the career paths of numerous researchers at different stages of their careers were discussed. Citizen scientists were also invited to a “Kefir Day,” for which they were introduced to the research and laboratory facilities within the Teagasc Food Research Centre, Moorepark. Finally, completion certificates, Kefir4All merchandise, and an individualized report detailing and graphically displaying the microorganisms in the participant’s kefir (Fig. 5; Supplementary Reports 1 and 2 were highlighted as materials that would be provided at the end of the project).

We noted the consistent number of samples provided by regular fermenters, across all stages of the project and a subtle increase in the number of samples provided by regular fermenters at week 09 and week 17, which corresponded to the introduction of both the engagement activities described (Fig. 3D). Furthermore, according to the Evaluation Survey, the majority of citizen scientists and school coordinators reported that the engagement activities contributed to continued participation (see below for further details).

The chi-square test of independence was employed to analyze the relationship between the categorical variables in the Fermentation Survey and fermentation characteristics such as grain growth, pH values, and the frequency of the observed “Fermentation characteristics” (Fig. 6). Categorical variables included “Fermentation categories,” which, in the case of milk kefir included “Full fat milk,” “Low fat milk,” “Goat milk,” and “Skimmed milk.” “Fermentation characteristics” in milk kefir included “Bubbling,” “Coagulation,” “Whey and curds,” and “Unchanged.” For water kefir, categorical variables included “Fermentation categories” such as “White sugar and one dried fig,” “White sugar, one dried fig, 1 slice of fresh lemon,” “White sugar and one dried apricot,” and “Brown sugar.” The “Fermentation characteristics” in water kefir included “Biofilm,” “Bubbling,” and “Unchanged’”(Fig. 6A).

Following P-value adjustment for multiple comparisons, the categorical variables “Fermentation categories,” “Grain wash,” “Room temperature,” “pH,” “Grain weight,” and “Liquid volume” were all shown to be significantly associated with the observed fermentation characteristics (see Materials and Methods) in milk kefir (Pearson’s chi-squared test, adjusted P-values <0.01). Similarly, the categorical variables “Fermentation categories,” “Grain wash,” “Room temperature,” “pH,” “Grain weight,” “Water type,” “Liquid volume,” and “Sugar weight” were all shown to be significantly associated with the observed fermentation characteristics in water kefir (Pearson’s chi-squared test, adjusted P-values <0.01). The strongest positive correlations were primarily observed between the “Fermentation categories” and “pH” values for both milk and water kefir when considering “Fermentation characteristics.” “Liquid volume” and “Grain weight” demonstrated a number of strong correlations in the milk kefir cohort (Fig. 6B). Notably, for water kefir, the fermentation parameters “Liquid volume” and “Sugar weight” contained a limited range of values as participants largely adhered to the recommended set ratios of 300 mL of water to 12 g of water kefir grains during the study (Fig. 1). However, the 300 mL of milk to 12 g of milk kefir grains was not as readily adopted due to difficulties in grain growth throughout the project.

Within the closing comments of the fermentation survey, grain was the word that most frequently appeared (16 participants in the water kefir cohort and 11 participants in the milk kefir cohort). Closer inspection highlighted that six participants in the milk kefir cohort and 11 participants in the water kefir cohort expressed concerns over grain growth, for example, “I’m quite concerned at the fact that they didn’t grow.” Another common theme in relation to the grain-specific comments from citizen scientists was a change in the appearance of the kefir grains with five water kefir participants and two milk kefir participants reporting physical deviations in their grains.

As a consequence of the recruitment and engagement initiatives undertaken during the Kefir4All project, 72 participants provided 516 milk and water kefir grains and liquid samples (Fig. 3D and 7C). These 516 samples were further divided into 223 milk kefir samples from 29 participants and 293 water kefir samples from 43 water kefir participants. Thanks to the contributions of citizen scientists, Kefir4All provided a unique opportunity to study natural changes in the milk and water kefir microbiome within household settings and in response to different fermentation parameters while also facilitating a comparison of the species-level data with that from other kefir studies employing a more traditional method of sample generation and collection (Fig. 7). After DNA extraction and high-throughput shotgun metagenomic sequencing, the microbiome composition of milk and water kefir samples collected in Kefir4All was determined and compared with that of the 256 milk (17) and 375 water kefir microbial profiles (unpublished data) (Fig. 7B and C), produced under controlled laboratory settings. The milk and water kefirs produced under controlled laboratory settings were initiated using 64 distinct milk kefir grains and 69 distinct water kefir grains, respectively. These grains were originally sourced from >20 different countries, and repeat 24–48-hour fermentations for each kefir grain sourced are represented in Fig. 7A (Y-axis). Milk and water kefir profiles generated by the Kefir4All citizen scientists initiated using single-source microbial communities mirrored the diversity observed in both milk and water kefir profiles initiated using different grains (Fig. 7A and B) with the added advantage of associated metadata in the form of 1,823 fermentation surveys collected over 21 weeks (Fig. 3C) (see Materials and Methods). Specifically, hierarchical clustering analysis identified eight community-type clusters across the study groups, with “Breselge et al.—water kefir,” “Kefir4All—water kefir,” “Kefir4All—milk kefir,” and “Walsh et al.—milk kefir” metagenomes representing the data sources considered in this study (Fig. 7B). A value of 0.1 Dunn index score was calculated for this clustering analysis; Jaccard coefficient values obtained using bootstrapping were used to evaluate cluster stability, and values indicated the frequency in which metagenomes of each cluster grouped together. Values ranged from 0.46 to 0.91; values <0.60 were considered unstable, and values >0.85 were considered highly stable. Of the eight clusters identified (Fig. 7B), clusters 1, 2, and 3 were considered unstable. Clusters 1, 3, 4, and 5 contained metagenomes from Breselge et al.—water kefir and metagenomes from Kefir4All—water kefir. Cluster 2 contained metagenomes from all data sources but most frequently from water kefir-specific data sources but also contained the lowest Jaccard coefficient of 0.46. Clusters 6, 7, and 8 contained metagenomes from Walsh et al.—milk kefir and Kefir4All—Milk kefir (Fig. 7B). As expected, more representative metagenomes for each cluster were identified in Walsh et al.—milk kefir and Breselge et al.—water kefir, with notable exceptions such as cluster 4, containing 15 metagenomes from Breselge et al.—water kefir and 69 metagenomes from Kefir4All—water kefir and cluster 7 containing 23 metagenomes from Walsh et al.—milk kefir and 109 metagenomes from Kefir4All—milk kefir.

As noted, the citizen scientists were informed that, at the end of the study, they would receive a report detailing the microbial composition of the kefirs they produced, including in the microbial profile over the course of the project (Fig. 5C; Supplementary Reports 1 and 2). Prior to circulation of the personalized reports, a number of citizen scientists were asked to review the report and provide feedback. Comments included “My overall impression is that it is an excellent individual Report tailor made for me. The language of the Report, I felt, was geared towards citizen scientists and therefore easy to understand.” Within this report, appropriate reading material was provided, to ensure participants had the necessary scientific knowledge to interpret the data. The appropriate reading material conveyed background information concerning microbiology, fermented foods, the microbial communities contained within milk kefir and water kefir (Fig. 5C), and the different measures available for the description of milk and water kefir microbial communities, including taxonomy, relative abundance, and alpha and beta diversity. Participants from whom no samples were received were provided with a reference report to ensure they were informed about the overall outcomes of the study (Supplementary Reports 1 and 2)

The Evaluation Survey (Fig. 8) completed by both citizen scientists (n = 18) and supervisory teachers (n = 6) revealed that all responding citizen scientists and teachers felt that their interest in science (Fig. 8A through D, G, H, O, and P) and understanding of scientific vocabulary relating to fermented foods or microbiology (Fig. 8E and F) were increased as a result of taking part in the project. Specifically, 10 (42%) of the responders, including both supervisory teachers and citizen scientists, felt that they or their students had moderate improvements, while seven responders reported a large improvement in their understanding of scientific vocabulary (Fig. 8E and F). Additionally, both teachers and citizen scientists felt that their general knowledge of microbiology and/or fermented foods, or those of their students were improved after taking part in the project. Specifically, none of the 24 responders (0%) reported that they did not benefit in terms of knowledge and interest of microbiology and fermented foods at the end of the project, and 17 responding citizen scientists (94%) and 4 teachers (67%) reported improvements in knowledge and interest in both microbiology and fermented foods (Fig. 8C and D). Lastly, 16 of the 18 responding citizen scientists (89%) and all 6 teachers (100%) expressed an interest in participating in future citizen science studies (Fig. 8O and P).

When analyzing the individualized responses to the questions “Why did you proceed with or stop the project?,” the most frequently occurring word was “kefir,” which was referenced in 6 of the 18 responses (33%) and further referred to as “it” or “fermented food” in 4 additional comments (22%) (Fig. S2). Full comments included “I was interested in observing changes in the kefir from week to week,” “I was very curious to see how the microbes in my kefir changed over the course of the project,” and “I enjoyed doing it, and it was fun to see what would happen.” “Interested” and “interesting” were the next most frequent words, which were referenced in 5 (28%) and 3 (16%) of the 18 responses, respectively, and concerned the areas of microbiology and food fermentation (Fig. S2). We observed identical trends when analyzing the individualized comments to the question “What did you like about the project?,” “kefir” was again the most frequent word accounting for 8 of the total 24 responses (33%), followed closely by liked (n = 7, 29%), interesting (n = 5, 21%), different (n = 5, 21%), and seeing (n = 4) (Fig. S3 and S4). Full comments included “I learned more about kefir and good conditions for growing it,” “I liked seeing different results after each fermenting,” and “That students could complete it at home, and they were expected to be responsible.” Engagement activities helped to maintain interest in the responding participants, with just one participant recording that the engagement activities did not maintain their interest. The majority of responders, including both teachers (n = 4, 67%) and citizen scientists (n = 11, 61%), felt that all of the engagement activities combined assisted in maintaining interest (Fig. 7K and L), while others, including both teachers and citizen scientists, highlighted the importance of the Kefir Day (n = 3, 12%) or the provision of Kefir4All merchandise (n = 3, 12%) (Fig. 7M and N). “Nothing” was the most frequent word included in the personalized responses to the question “What did you dislike about the project?” (Fig. S5 and S6), accounting for 7 of the total 24 responses (29%), with no other general trend of interest occurring in terms of word frequency (Fig. S7 to S10).