Computational Analysis of the Effect of Dietary Interventions on the Gut Microbiome Composition in Parkinson’s Disease

Abstract

Parkinson’s disease (PD) is a neurodegenerative disorder characterized by motor symptoms like tremor, rigidity, and bradykinesia. The WHO estimates that 10 million people currently have PD, with its prevalence expected to double to 20 million by 2050. Key risk factors include age, male sex, environmental contaminants, and family history. Emerging evidence links gut microbiota dysbiosis to PD, suggesting it contributes to neuroinflammation and disease progression, though the role of dietary interventions remains unclear. This study used computational simulations with genome-scale metabolic models (GEMs) to analyze how diet impacts the gut microbiota in PD patients. Fecal microbiota from PD patients and healthy controls were compared across three diets: high-fiber, Mediterranean, and vegan. Simulations revealed increased pro-inflammatory bacteria (e.g., Escherichia coli O157) in PD patients, likely due to reduced bacterial competition, alongside the decreased production of beneficial metabolites like butyrate, phenylalanine, and cysteine. The Mediterranean diet showed higher short-chain fatty acid production, potentially benefiting PD patients. These findings underscore the importance of dietary interventions in modulating the gut microbiome and suggest that targeted diets may complement PD therapies, improving patient outcomes.

1. Introduction

Parkinson’s disease (PD) is a neurodegenerative disorder marked by the progressive loss of dopaminergic neurons in the substantia nigra, driven by a combination of factors including α-synuclein aggregation, mitochondrial dysfunction, neuroinflammation, impaired protein clearance, genetic mutations (e.g., LRRK2, PINK1, PARKIN), and oxidative stress from dopamine metabolism [1]. PD arises from a complex interplay of genetic, environmental, and lifestyle factors, with the gut microbiome emerging as a key contributor [2,3,4,5]. Altered gut microbiota composition and diversity are consistently observed in PD patients, potentially contributing to the disease via neuroinflammation, oxidative stress, and neuroactive metabolite production [4,5].

PD-associated gut dysbiosis imbalances are characterized by both an overrepresentation of inflammatory/opportunistic pathogens and a loss of anti-inflammatory commensals, leading to altered microbial community function [6]. Studies have highlighted microbial changes linked to neurotransmitter synthesis and inflammation [7], increased intestinal permeability [8], and specific bacterial genera associated with clinical phenotypes [5]. Microbial signatures have also been correlated with motor symptom severity [9], suggesting a potential for the gut microbiome as a biomarker and therapeutic target.

Metabolic findings, however, remain inconsistent. While [9] reported lower acetate, butyrate, and propionate in PD fecal samples, others found only reduced acetate and butyrate [10], with no plasma SCFA differences [11]. This underscores the need for advanced methodologies to elucidate the microbiota–host relationship [12]. In this regard, genome-scale metabolic modeling (GEMS) has advanced the understanding of the microbiota’s impact on health [12,13,14]. For instance, recently we analyzed gut microbiomes in 25 PD patients and 25 healthy controls, revealing altered bacterial taxa and metabolite production in PD [3].

This research delves deeply into the intricate relationship between gut bacteria, diet, and metabolic potential in PD. By meticulously selecting and analyzing data from two highly regarded microbiome studies in PD using advanced computational methods, we aimed to uncover possible transformative dietary strategies for effective disease management.

2. Materials and Methods

2.1. Selection of Studies for Simulation

From an initial list of 6 studies characterizing the gut microbiome in PD [5,8,15,16,17,18] the studies by [8,15] were selected for this study, since these studies reported bacterial abundance percentages at a high taxonomic resolution even when considering low-abundance groups, they had bacterial species records, and their study methodology was similar enough to be comparable. Subsequently, for the Keshavarzian study, a detailed list of the distribution of bacterial genera in PD and healthy controls (HCs) was generated (Supplementary Table S1). For the Bedarf study, a detailed list of the distribution of bacterial families in PD and HCs was generated (Supplementary Table S2).

Selection of Metabolic Models for Microbiome Reconstruction

The bacterial metabolic models for the simulation were obtained from Virtual Metabolic Human [19], according to the following criteria.

For the study by Bedarf et al. [15], the species-level metabolic reconstruction with the lowest number of blocked reactions (RBs) was selected for each family (Table S3). For the study by Keshavarzian et al. [8], the species-level reconstruction with the lowest RBs was selected for each genus through the g2f package in R [20] (Table S4). All selected models were individually revised and optimized for biomass production.

Subsequently, through the sybil package [21], the production of short-chain fatty acids (acetate, butyrate, and propionate), in bacterial compositions in HCs and PD, was evaluated according to bacterial rates of SCFA production [9]. Thus, reconstitutions with acetate, butyrate, and propionate synthesis reactions were employed. These selection filters ensured lower computational costs and homogeneity in the metabolic models under study.

2.2. Competition and Complementarity Assembly

Following the protocol given by [22], which states that bacterial relationships are determined by competition (two bacterial species use the same metabolites) and by complementarity (a bacterial species A can employ metabolites produced by a species B), we established the degree of overlap in the metabolites consumed by two species according to (1) as follows:

MCI(A,B) = |CA CB| min(|CA|, |CB|)

(1)

where CA and CB are the sets of compounds consumed by species A and B, respectively.

The complementarity index measures how well the metabolites produced by one species can be used by another, according to (2) as follows:

MComI(A,B) = |PA CB| + |PB CA|min(|PA| + |CB|, |PB| + |CA|)

(2)

where PA and PB are the sets of compounds produced by species A and B, respectively.

Complementarity and competition indice values range from 0 (minimal) to 1 (maximal) for competition or complementarity.

Therefore, for both bacterial community reconstructions, models were developed to evaluate competition for metabolite resources and complementarity in biomass stimulations. Complementarity relationships were specifically examined to establish interactions between SCFA producers and other members of the bacterial community. Additionally, the complete list of secreted and consumed metabolites was obtained for each reconstruction model.

2.3. Generation of a Computational Model of the Bacterial Composition of the Colon

Based on taxonomic information and curated computational models, computational simulations were performed using the BacArena package [23]. These compositions were evaluated in three different metabolic scenarios according to three types of diet, Mediterranean, high-fiber, and vegan (Supplementary file Table S5), where metabolites differed in flux value (mmol/DW/h) [19]. Fifty replicates were performed for each combination, diet, and composition of HCs and PD.

The simulations were performed under anaerobic conditions [24]. Over a time of 8 h, the average filling time [25,26] and the time it takes for the chyle to traverse the colon [27] was simulated, on a 100 × 100 grid, according to a standard protocol [23,28,29]. Modeling was initiated with a bacterial inoculum of 250 cells per HC and PD composition, as this inoculum fully occupied the available space within the simulation period. The proportion in the abundance of each bacterial group had the highest taxonomic resolution, obtained in the chosen studies.

2.4. Availability of Bacterial Metabolites to the Host

The concentration of metabolites related to amino acid networks, carbohydrates, short-chain fatty acids, gases, and vitamins were retrieved from the results at the end of each simulation. To obtain the metabolic availability (DMi) for the host in each context, the total bacterial consumption of metabolite i (ci) was subtracted from the total bacterial metabolite production (Pi) (3) as follows:

DMi = Pi − Ci

(3)

2.5. Topological Analysis of Networks

In order to establish the relationships between key metabolite producers, a topological analysis of the bacterial carbohydrate and amino acids networks was performed for PD and HCs under each of the three different diet scenarios, inferring standard centrality measures. For this, the results of the BacArena simulations were modified using igraph [30] and the exchange of amino acids and carbohydrates between all bacteria in PD and HCs was obtained from a global exchange network where nodes represented individual bacteria and edges the interchanged particular amino acid. Standard topological centrality measures were then obtained by means of igraph.

2.6. Statistical Analysis

The data obtained from the simulation were subjected to the Kolmogorov–Smirnov test and Bartlett’s test to verify their normality and homogeneity of variance, respectively. In addition, the Wilcoxon univariate test was applied to identify differences in metabolite production and consumption among the different diets. To control the error associated with multiple comparisons, the False Discovery Rate (FDR) was adjusted using the Benjamini–Hochberg procedure.

Tests were performed to verify the normality and homogeneity of the residuals, as well as autocorrelation and multicollinearity analyses to ensure the necessary assumptions for the application of multiple regression. Since the results obtained for the relationship between the abundance of pro-inflammatory bacteria and their complementary and competitor bacteria do not follow a linear relationship, the evaluation was performed using the LOESS regression method. To evaluate the contribution of each of the diets in the production of metabolites of interest such as acetate, butyrate, and propionate, a principal component analysis was performed.

3. Results

3.1. Competition Indices Outweighed Complementarity in Bacterial Communities

Since bacterial diversity appeared to be similar between PD and HCs in both studies under analysis, competitiveness and complementary indexes were determined for all samples in both studies.

For the study by [15], the average competition index was 0.249 ± 0.2, while complementarity averaged at 0.085 ± 0.02. Notably, Bifidobacterium_longum_E18 exhibited the highest competition index with an average value of 0.367 (Figure 1A). Among bacterial pairs (Supplementary Table S6), Dorea formicigenerans str. ATCC 27755 competed strongly against Faecalibacterium prausnitzii str. M21 2 (index = 0.667), and Ruminococcus torques L2 14 also competed against F. prausnitzii str. M21 2 (index = 0.651). Furthermore, EHL27 exhibited the highest average complementarity index in the Bedarf study, averaging 0.132 (Figure 1B). The most complementary pairs (Supplementary Table S6) were Bacteroides stercoris str. ATCC 43183 and Escherichia coli str. SE11 with an index of 0.16. In the Keshavarzian et al. [8] study, the average competition index was 0.236 ± 0.22, and complementarity averaged at 0.06 ± 0.03 (Supplementary Table S7). The species with the highest average competition index (Figure 1A) was Blautia hanseii VPI C7-24 DSM 20583 with an index of 0.331. The most competitive pairs (Supplementary Table S7) included Roseburia hominis str. A1 183 and Eubacterium hallii str. DSM 3353 (competition index = 0.583), Bifidobacterium catenulatum str. DSM 16992 and Bifidobacterium angulatum str. DSM 20098 (0.727), and Parabacteroides sp. str. D13 with Bacteroides sp. str. 217 (0.571). Conversely, the highest complementarity coefficients (Figure 1B) were observed in the species Bacteroides sp. 4_3_47FA (BS4FA) with an index of 0.139 and in the bacterial pairs (Table S7) such as Bacteroides sp. 4 2 47AA with Escherichia_sp_3_2_53FAA (0.177), Bacteroides sp. 4 3 47FAA with Bacteroides sp. 217 (0.177), and Bacteroides sp. 4 2 47AA with Prevotella stercorea str. DSM 18206 (0.168).

Figure 1. Species relationship heatmap of the bacterial communities modeled in the Bedarf et al. [15] and Keshavarzian et al., [8] studies for PD. (A): Competitiveness indexes. (B): Complementarity indexes. The list of abbreviations used in this graph is found in Table S8.

3.2. In the Parkinsonian Context, Pro-Inflammatory Bacteria Are Opportunistic

Our results suggest that, in PD, the proliferation of pro-inflammatory bacteria is associated with the abundance of their competitor and complementary bacteria, being statistically significant in both cases (p < 0.01). In this scenario, for instance, the abundance of E. coli O157 H7 str. Sakai in [15] is mainly associated with its competitive bacteria E. coli SE11 (Supplementary Tables S6 and S11) not exerting an effective antagonism effect (0.123) (Figure 2A). On the contrary, it can be noted that in HCs, competition exerts regulatory effects on the abundance of E. coli O157 H7 str. Sakai by limiting its proliferation (0.453). For the case study in Keshavarzian et al. [8], a similar case was observed: Escherichia sp. 3 2 53FAA was regulated in HCs by the competitive bacteria Escherichia sp. 4 1 40B whereas, in PD the complementary bacterium stimulated the growth of Escherichia sp. 3 2 53FAA (Figure 2B).

Figure 2. Regression of pro-inflammatory bacteria. (A) Bedarf et al. [15] study. (B) Keshavarzian et al. [8] study.

The abundance of bacteria involved in these interactions varied in the contexts evaluated. In [15], the abundance of E. coli O157 H7 str. Sakai was significantly different (p-value < 0.05, Wilcoxon’s test; FDR-adjusted p = 0.041) between PD and HCs, being higher in PD (mean = 164; median = 179) than in HCs (mean = 149; median = 168). Its complementary bacterium, F. prausnitzii M21 2, showed a slightly lower abundance in PD (mean = 404; median = 227) compared with HCs (mean = 423; median = 241), being also statistically different (p-value < 0.05, Wilcoxon’s test; FDR-adjusted p = 0.035). On the other hand, its competitive bacteria E. coli SE11 was less abundant in PD (mean = 129; median = 141) than in HCs (mean = 176; median = 205), with a statistically significant difference (p-value < 0.05, Wilcoxon’s test; FDR-adjusted p = 0.021).

A similar result was observed in the study by [8]. In this study, the pro-inflammatory bacterium Escherichia sp. 3 2 53FAA was more abundant in PD (mean = 61; median = 131) than in HCs (mean = 30; median = 137), with this difference being statistically significant (p-value < 0.05, Wilcoxon’s test; FDR-adjusted p = 0.043). Its complementary bacterium was also less abundant in PD (mean = 519; median = 658) than in HCs (mean = 849; median = 718). While its competitor Escherichia sp. 4 1 40B was statistically (p-value < 0.05, Wilcoxon’s test; FDR-adjusted p = 0.045) less abundant in PD (mean = 27; median = 137) and more abundant in HCs (mean = 56; median = 131).

3.3. Metabolic Exchange Networks Leave Certain Metabolites Less Available in PD

Our model suggests that there are differences between the host availability of key metabolic products between HCs and PD. In this sense, metabolites such as aspartic acid, alanine, cysteine, leucine, serine, phenylalanine, glucose, maltose, lactate, and riboflavin showed statistical differences (p-value < 0.05, Wilcoxon’s test; FDR-adjusted p = 0.029). Overall PD was shown to be the most affected phenotype, showing significant reductions in most bacterial metabolic products (Figure 3).

Figure 3. Boxplot of common metabolite concentrations in the available analyzed studies. Red box: control, Bedarf et al. [15]. Green box: control, [8]. Blue box: Parkinson’s, Bedarf et al. [15]. Purple box: Parkinson’s, [8]).

3.4. Changes in PD in the Individual Amino Acid Production Network

Given the importance of tryptophan availability in PD, we focused part of our analysis on this particular network. Our results suggest that in the two studies evaluated tryptophan had a lower number of producers (outdegree) in PD, which is more evident in the study by Keshavarzian et al. [8] (Figure 4, bottom right) and that the number of consumers (indegree) was the same in HCs and PD (Figure 4). In addition, the high-fiber diet had one less phenylalanine producer in PD in both studies. However, some variations in the producers of certain amino acids were unique to each study.

Figure 4. Tryptophan network. (A) Bedarf et al. [15]. (B) Keshavarzian et al. [8]. Datasets. Nodes represent bacterial taxa; node color indicates indegree (number of metabolite consumers) and node size indicates outdegree (number of producers). The network shows fewer tryptophan-producing bacteria in Parkinson’s disease (PD) than in healthy controls (HC), particularly in Keshavarzian et al. [8], highlighting reduced metabolic capacity in PD. Left: HC. Right: PD.

Regarding other amino acids, in the case of the study by [8] the Mediterranean diet presented one more phenylalanine producer in PD (n = 5) than in HCs (n = 4). The amino acid lysine had one more producer in the high-fiber and vegan diets (n = 6) and one less (n = 5) in the Mediterranean diet. However, in the context of PD all diets had five producers of this amino acid. The Mediterranean and vegan diets had one more producer (n = 4) of cysteine compared with the high-fiber diet (n = 3). For the case of the Bedarf et al. [15] study, in the vegan diet in PD, one more producer (n = 2) of isoleucine was quantified (Table S12).

3.5. Butyrate Dynamics Mediated by Diet

Bacterial SCFA concentrations were different between HC and PD. For the Bedarf et al. [15] study overall, acetate, butyrate, and propionate concentrations were higher in HCs than in PD (p-value < 0.05) (Figure 5A). In [8], acetate and propionate concentrations were higher in PD than HCs (p-value < 0.05), while butyrate concentrations were lower in PD (p-value < 0.05) (Figure 5B).

Figure 5. Acetate, butyrate, and propionate concentrations. (A) Study by [15]. (B) Study by [8].

The bacterial composition of PD shows a particular behavior towards butyrate. In general, it was produced less and consumed in higher concentrations in PD.

In the study by [15], in PD, butyrate was produced less in high-fiber and vegan diets (p-value < 0.05). In the study by [8], all diets presented lower butyrate production in the disease (p-value < 0.05) (Table 1). It can be highlighted that diets high in fiber produced the highest butyrate concentrations.

Table 1. Average butyrate production in the different diets (mmol).

Study	Diet	Healthy Controls (mmol)	Parkinson’s (mmol)	p-Value
Bedarf	Fiber	4.12003 ± 1.5456	3.97040 ± 1.2456	0.03
	Mediterranean	4.111063 ± 1.4450	4.037472 ± 1.3235	0.423
	Vegan	4.107646 ± 1.5344	3.888306 ± 1.6568	0.006
Kashavarzian	Fiber	8.81435 ± 2.2341	8.319964 ± 2.1324	2.02 × 10−42
	Mediterranean	8.628139 ± 2.1568	3.760015 ± 2.2756	1.68 × 10−45
	Vegan	8.713214 ± 2.2341	3.767233 ± 2.3876	6.23 × 10−46

Regarding butyrate intake, it can be highlighted that, for the high-fiber and vegan diets of [15] along with the three diets of [8], it was higher in the PD context. For the Mediterranean diet from the Bedarf et al. [15] study, the intakes were statistically similar (p-value > 0.05) in HCs and PD (Table 2).

Table 2. Average consumption of butyrate in the different diets (mmol).

Study	Diet	Healthy Controls (mmol)	Parkinson’s (mmol)	p-Value
Bedarf	Fiber	5.571× 10−9 ± 1.2× 10−5	2.449273× 10−9 ± 1.2× 10−4	0.007
	Mediterranean	1.0356× 10−9 ± 1.2× 10−4	1.821× 10−9 ± 1.2× 10−4	0.488
	Vegan	4.225× 10−10 ± 1.3× 10−5	1.391× 10−9 ± 1.2× 10−4	3.605 × 10−5
Kashavarzian	Fiber	3.440652 ± 5.412	7.250732 ± 5.018	2.036 × 10−28
	Mediterranean	3.363205 ± 6.324	6.495293 ± 6.578	2.506 × 10−31
	Vegan	3.481832 ± 8.951	6.495293 ± 5.610	3.304 × 10−28

3.6. In Search of a Healthy Diet for PD, Specifically for AGCC

To assess the individual contribution of diet to the promotion of healthy metabolites, PCA was performed in relation to the production of SCFAs and other metabolites. The component analysis obtained a total of 11 components for the PD cases in the Bedarf et al. [15] study. The sum of the first two components explains 85.76% of the cumulative variance (Table S9). In this regard, component 1 largely explains the production of SCFAs (acetate, butyrate, and propionate), indole, and tryptophan. Figure 6 shows two groups of metabolites that could be related. The first consists of acetate, formic acid, ammonium, and carbon dioxide, clustered around the vegan diet data points, while the second consists of butyrate, propionate, and tryptophan which are closer to the Mediterranean and high-fiber diets.

Figure 6. Principal component analysis (PCA) of metabolite production simulated for the Bedarf et al. [15] dataset. The first two components explain 85.76 % of the variance, separating diets by their metabolic profiles. Vegan diets cluster with acetate, formic acid, and ammonium, while Mediterranean and high-fiber diets associate with butyrate, propionate, and tryptophan, suggesting these diets favor beneficial metabolite synthesis.

For the results of the diets in [8], the component analysis found two components explaining 76.8% of the variance (Table S10). The first one is supported by acetate, butyrate, propionate, and glucose concentrations. The second explains the concentrations of guanosine and hydrogen. It is observed that, according to the distribution of metabolites in the plane, propionate, acetate, and butyrate could be correlated. At the same time, it is detailed that the latter are probably more related to Mediterranean and vegan diets (Figure 7).

Figure 7. Principal component analysis (PCA) of metabolite production simulated for the Keshavarzian et al. [8] dataset. Two principal components explain 76.8 % of the variance. Acetate, propionate, and butyrate concentrations are closely correlated and linked to Mediterranean and vegan diets, indicating that these dietary patterns promote short-chain fatty acid (SCFA) production. CA plots for [8].

4. Discussion

Gut bacterial composition plays a crucial role in PD, and in this study we set out to understand the relationship between healthy patients and those with the disease. Our initial findings reveal that in PD, bacterial communities show a higher rate of competition than complementarity, which is a result also observed in similar research. The prevalence of high competition values impacts community dynamics, as the presence of highly competitive members reduces resilience to physicochemical stressors. In contrast, highly complementary bacterial communities function as interconnected modules that can maintain a stable composition even under stress [31]. This particular dynamic in PD patients is related to the coefficient of variation results obtained in this study. The high value obtained indicates that the bacterial community in PD is more prone to undergo new compositional changes when experiencing stress scenarios, as its resilience and functional redundancy capacities are lower than those of HCs [32].

Regarding population dynamics in gut microbiota, it was found that pro-inflammatory bacteria abundance such as E. coli O157:H7 str. Sakai or Escherichia sp. 3 2 53FAA is promoted in PD patients, which seems to be favored by the absence of control exerted by competition, given by other noninflammatory bacteria belonging to the Escherichia genus such as E. coli SE11 and Escherichia sp. 4 1 40B, respectively. This fact could suggest an involvement in the development of PD through LPSs, recognized endotoxins derived from Gram-negative bacteria that activate microglia and release neurotoxic factors contributing to dopaminergic neurodegeneration [33]. Likewise, E. coli outbreaks in mice have been linked to the development of intestinal inflammatory conditions mediated by the O-antigen of its LPSs [34]. Furthermore, ref. [5] reported, through the fecal microbiota analysis of 72 PD patients and 72 controls, that in PD patients the abundance of pro-inflammatory bacteria is related to the association behavior between bacterial groups.

In addition, E. coli O157:H7 is an enterohemorrhagic serotype (EHEC) capable of producing the Shiga toxin (STEC), with some strains harboring the plasmid pO157, further increasing pathogenicity [35,36]. In humans, this strain causes acute gastroenteritis, hemorrhagic colitis, and hemolytic–uremic syndrome (HUS) [36], and has been reported in several outbreaks [37], underscoring its pro-inflammatory capacity. Strain-specific metabolic models have also shown differential auxotrophies for vitamins such as niacin (B3), thiamin (B1), and folate (B9) [37]. These metabolic distinctions support the idea that pathogenic EHEC strains may benefit from specific substrates, such as sucrose, which promote their growth [37].

Network analysis further suggested a decreased availability to the host of metabolites associated with PD development, including phenylalanine, cysteine, riboflavin, glucose, and leucine. Phenylalanine is the precursor of tyrosine and L-DOPA, with the latter being the direct precursor of dopamine [38,39]. In PD, dopamine deficiency is closely associated with motor symptoms [39], and a decrease in circulating phenylalanine has been documented [40]. Cysteine plays a controversial role: while high levels may contribute to neurodegeneration through NMDA receptor activation [41], N-acetylcysteine has also been proposed as a therapeutic strategy [42]. Riboflavin deficiencies have been observed in PD patients [43], and supplementation has been associated with reduced oxidative stress and motor improvements [43,44].

Short-chain fatty acids (SCFAs) play a crucial role in maintaining host health. Among these, butyrate is particularly known for those beneficial effects. In simulations conducted in the context of PD, a deficiency in butyrate was observed, which may have implications for health. It has been shown that a low butyrate concentration in PD minimizes the amount of SCFAs available to colonocytes [9], affecting their uptake as an energy source [45]. It has also been shown that a reduction in butyrate concentration results in an increase between 30 and 70% of oxidative phosphorylation levels [46]. To overcome this problem, ref. [47] demonstrated that the addition of sodium butyrate (200 mg/K) reduced both neurological oxidative effects in mice with traumatic brain injury and blood–brain barrier damage through the increased expression of occludins and ZO-1 (tight junction proteins) proteins [47]. In addition, sodium butyrate treatment has been shown to significantly reduce the effects of PD, acting as a neuroprotectant in mice with PD induced by alpha-synuclein aggregation (α-syn) [48]. It was also demonstrated that sodium butyrate has protective effects against the cytotoxic effects of 1-methyl-4-phenylpyridinium (MPP+), preserving the integrity of the membrane and neurites in LUHMES cells (human embryonic neural precursor cells) [49]. It has also been identified that, through interactions with GPR109A (G protein-coupled receptor, present in epithelial, dendritic, and macrophage cells) and mucin-2, sodium butyrate can suppress inflammation and maintain intestinal permeability integrity in mice, respectively [50,51].

By analyzing the three dietary scenarios, we observed that two of them (high-fiber and Mediterranean) were associated with SCFA production, suggesting that these diets may promote the microbial synthesis of beneficial metabolites. Soluble fiber intake has also been found to form the basis of fermentative processes used for the microbial synthesis of acetate, butyrate, and propionate [52]. This is in agreement with studies such as [53], who found that in elderly people aged 76–95 years, fiber consumption (both soluble and insoluble) is associated with an increase in SCFA concentrations. In addition, ref. [54] observed that patients exposed to soluble fiber intake after their time in the Intensive Care Unit (ICU) had elevated levels of SCFA-producing bacteria, which has anti-inflammatory effects on intestinal epithelial and immune cells [55].

These findings are consistent with the results of our principal component analysis (PCA), where high-fiber and Mediterranean diets were associated with acetate, butyrate, and propionate production. This link suggests that such diets may play a crucial role in modulating the gut microbiota and thus in the production of beneficial metabolites, which could have important implications for gut and systemic health, including possible symptom mitigation in PD patients.

It has been identified that adherence to this diet decreased the probability of the onset of prodromal Parkinson’s in patients aged 65 years or older [56], as well as being inversely associated with the onset of prodromal PD, constipation, daytime sleepiness, and long-term depression [57]. Likewise, a lower probability of PD onset has been reported in patients with higher adherence to the Mediterranean diet, while earlier disease onset is associated with lower adherence to this diet [58]. Finally, some phenolic compounds, carotenoids, and vitamins C and E can reduce inflammation in patients with PD [59], and early adherence to the Mediterranean diet has been observed to delay the onset of PD in men by approximately 8.4 years [60]. However, these findings should be interpreted as hypothesis-generating rather than prescriptive. While our simulations align with epidemiological evidence, they cannot substitute experimental validation, and further longitudinal studies are needed to confirm whether dietary interventions truly modulate PD onset or progression.

Finally, we acknowledge a limitation of our study. The reliance on Keshavarzian et al. [8] and Bedarf et al., [15] was driven by their taxonomic resolution and methodological comparability, which enabled reconstructions at the species and family levels. Nevertheless, expanding the dataset base in future work will be essential to capture additional variability and enhance the robustness and generalizability of the results.

5. Conclusions

The taxonomic clustering approach offers a useful in silico framework to explore the bacterial microbiota in PD, complementing wet-lab findings by simulating a limited set of species. Likewise, the reverse ecology analysis highlights conditions favoring the proliferation of pro-inflammatory bacteria; although, this method does not capture the full spectrum of metabolic interactions among taxa with distinct degradation and production capacities.

Our results suggest that dysbiosis in PD involves not only changes in abundance but also network-level disruptions, such as the loss of mutualistic interactions and producers of essential metabolites. These imbalances create a context where pro-inflammatory species can thrive, consistent with the increased sensitivity of PD microbiota to perturbations. In this setting, reduced levels of butyrate—a SCFA with specific neuroactive effects—appear relevant, as its demand exceeds bacterial production in PD patients.

Finally, dietary patterns may help modulate these dynamics. Our results suggest that the Mediterranean diet is associated with the increased bacterial production of short-chain fatty acids. In particular, this diet has been shown to increase the production of these beneficial metabolites in the two studies evaluated. While our simulations cannot be interpreted as clinical recommendations, they are consistent with epidemiological evidence suggesting that.