Prioritizing neglected food species in nutritional studies using expert-knowledge and explainable AI

Abstract

Food biodiversity is vital for human health and the development of sustainable food systems. However, research on neglected and underutilized species is limited by funding, uneven research capacity, and the challenge of balancing ecological, cultural, and public health considerations, requiring innovative prioritization approaches. Using Brazil as a model, this study inventories 369 neglected food species across algae, aquatic fauna, wild terrestrial vertebrates, insects, mushrooms, and plants. A mixed-methods approach, combining expert knowledge and explainable AI (LightGBM and SHAP value analysis), identified key factors for prioritizing species for food composition and consumption studies. The inventory is dominated by plants (29.5%) and wild vertebrates (24.4%), with significant gaps in nutritional data, particularly for algae, insects, and wild vertebrates. Over 36,000 recipes using neglected species were identified. In both food composition (R²: 0.677) and consumption studies (R²: 0.782), recipe number and species occurrence across different states were the most influential features in predicting prioritization. These findings emphasize the role of cultural uses and local accessibility in shaping nutritional research priorities. We urge increased research on neglected species to bridge data gaps and integrate them into food systems, promoting sustainable diets in Brazil and other tropical regions.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-39484-6.

Introduction

Food biodiversity, defined as the diversity of plants, animals, and other organisms used for food, both cultivated and wild, plays a crucial role in global health and food security¹. Within this context, neglected species, which include a range of wild and/or underutilized plants, fungi, algae, and wild animals (in specific contexts), represent an untapped resource for improving nutrition and sustainable food systems². While no objective criteria definitively separate neglected from non-neglected food species, a consensus defines neglected species as those whose contributions to sustainable food systems are significantly undervalued^3,4. This undervaluation stems from a widespread lack of awareness, limited information, and minimal presence in established markets.

The relevance of food biodiversity to human health and food and nutrition security (FNS) is multifaceted. Biodiversity significantly influences food availability, quality, and stability—key components of FNS that directly impact human health outcomes⁵. In fact, increased Dietary Species Richness (DSR) improves micronutrient intake adequacy⁶ and is associated with lower mortality rates, as shown by a large longitudinal cohort study conducted across nine countries⁷. Specifically, a 10-species increase in DSR was associated with a 14% to 17% reduction in all-cause mortality in men and a 6% to 8% reduction in women after an average 20-year follow-up.

These studies highlighted the association between higher dietary diversity and reduced mortality through four mechanisms⁷. The first, termed as “sampling effect”, posits that increasing DSR enhances the likelihood of including a wider array of highly nutritious and protective foods. The second, known as the “complementary effect”, suggests that interactions between different food species can result in health benefits greater than the sum of their individual effects. The third mechanism, "minimizing trade-offs", reduces the risk of toxicity from the overconsumption of any single species, even considering the natural presence of anti-nutritional factors or agrochemicals. Lastly, dietary diversity provides a broad range of nutrients and substrates, such as dietary fiber and bioactive compounds, that support the development of various gut microbes, fostering microbiome diversity. This change in gut microbiota, in turn, enhances health responsiveness, as a more diverse microbiome tends to be more stable and resilient to disturbances^8,9.

In this context, food composition and consumption studies are crucial for shaping nutritional science and public health policies. Food composition studies systematically analyze and document nutritional and chemical makeup of food, including macro- and micronutrients, bioactive compounds, and other essential components. The findings from these studies contribute to the creation and maintenance of food composition databases¹⁰, which are fundamental for establishing dietary guidelines and informing nutrition policies. On the other hand, food consumption studies collect detailed data on the types and amounts of foods consumed by specific populations, tracking dietary habits, consumption patterns, and factors influencing food choices¹¹. Therefore, food composition and consumption studies help identify the nutritional gaps and areas where dietary improvements are needed, making them indispensable tools in addressing malnutrition in all its forms and promoting public health. These two fields of study are deeply interconnected: food composition data is essential for accurately assessing nutrient intake, while food consumption data highlights which foods should be the focus of composition analyses¹².

Despite their importance, researchers face significant challenges in determining which neglected species should be prioritized for inclusion in food composition and consumption studies. Studies on food composition and consumption of neglected and underutilized food species face multifaceted challenges. The sheer diversity of species, uneven research capacity, and the complexity of integrating ecological, cultural, and nutritional priorities into a cohesive strategy complicate the decision of which species should be the target of research efforts. Prioritizing species for research requ1ires careful consideration of multiple factors, including nutritional value, cultural significance, ecological importance, and research needs, among others. These factors are exemplified in the literature and summarized in Box . These challenges highlight the urgent need for innovative, data-driven approaches to strategically prioritize species for study, ensuring that limited resources are effectively utilized to address critical gaps and unlock the potential of these species in advancing sustainable and healthy food systems.

Box 1.

Criteria for choosing neglected food species in nutritional studies.

Food Composition Studies 1. Research Gaps: Prioritize species with limited or outdated nutritional data12 2. Cultural Significance: Focus on culturally significant or traditionally consumed foods that lack comprehensive nutritional data49 3. Anecdotal or Preliminary Evidence: Consider species with anecdotal reports of health benefits or preliminary studies suggesting nutritional value that require further scientific examination50 4. Ecological and Agricultural Significance: Include species that play important roles in local ecosystems or agricultural systems but have yet to be thoroughly studied for their nutritional content51 5. Emerging or Underutilized Species: Prioritize species gaining popularity or those with potential for broader use that lack comprehensive nutritional information4 6. Biodiversity Conservation: Considers species at risk of genetic erosion or extinction to document their nutritional profile before potential loss, while simultaneously implementing protection measures and sustainable management strategies to prevent overexploitation resulting from increased awareness and demand52 7. Climate Resilience: Include species known for adaptability to changing environmental conditions to understand their nutritional profile53 Food Consumption 1. Survey goal: Species that reflect the objective of the proposed dietary assessment, which can be tracking specific food groups or nutrients, a specific health outcome, or a whole dietary pattern3 2. Local Availability and Accessibility: Focus on available species that are easily accessible to communities54 3. Cultural Significance: Prioritize culturally important or traditionally consumed species50 4. Underutilization: Include underutilized species to understand consumption patterns and the potential for promoting their intake5 Overlapping Criteria 1. Cultural Significance: This section emphasizes the importance of studying foods with traditional or cultural values in a community. Including such foods in studies helps preserve cultural knowledge and potentially uncovers nutritional content 2. Local Availability and Use: Prioritizes foods that are easily accessible to local communities. Studying locally available foods in both composition and consumption contexts ensures the research is relevant and applicable to the target population’s daily lives 3. Underutilized or Emerging Species: This category focuses on foods that are not widely used but have the potential for greater utilization. By studying these in both composition and consumption contexts, researchers can identify nutritionally valuable foods that could contribute to dietary diversity and food security 4. Research Goals and Gaps: This section emphasizes aligning the selection of foods with the overall objectives of the study, whether it is filling knowledge gaps in nutritional data or understanding consumption patterns

This article seeks to address the challenge of prioritizing neglected foods by focusing on two main objectives. First, we used Brazil as a model to compile and characterize an inventory of neglected food species in the country. Second, we utilized this inventory as a basis to develop and apply a data-driven framework for prioritizing food species based on nutritional composition and consumption studies, integrating expert knowledge and machine learning techniques. A thorough evaluation of the key factors driving species prioritization for nutritional composition and consumption studies can offer a replicable framework for guiding and informing effective resource allocation in both research and policy decision-making.

Results

Inventory of neglected food species

We cataloged 369 species, with plants (29.5%) and wild animals (24.4%) predominating (see Supplementary Information 1). Mushrooms represented 23.6%, followed by fish and seafood (13.3%), insects (6.8%), and algae (2.4%). Almost half of the species (46.9%) lacked information on their conservation status, either because they had not been evaluated (44.4%) or were classified as data deficient (2.4%). Among those assessed, 42.8% are classified as least concern, while a total of 10.3% of the species face some degree of threat, being classified as critically endangered (0.3%), endangered (1.1%), vulnerable (5.7%), or near threatened (3.3%). The majority of species (81.6%) are native to Brazil. As for the commercialization, only 28.9% of the species are commercially cultivated. A summary of the distribution of this species across the regions of Brazil, the relationship of these administrative regions with the country’s six biomes, as well as a highlight of unique species in each of these regions, is available in Fig. 1.

Fig. 1.

Distribution of 369 neglected food species across the administrative regions of Brazil. The map also depicts the distribution of Brazilian biomes, noting that all regions encompass more than one biome. Many food species occur in multiple regions, resulting in overlapping distributions. To emphasize region-specific patterns, unique species present in each region are highlighted. The North region, with the greatest overall and unique species richness, exemplifies the high biodiversity found in the country.

Source: Image created by the author with the Canva software (Canva Pty Ltd, 2024; https://www.canva.com).

Regarding food composition, just 33.1% of the species had available nutritional information (Supplementary Information 2). The group of foods that most frequently had nutritional information was plants, with 88.0% of the total species listed having available nutritional data, which is significantly higher than the nutritional information availability observed for other food groups, such as fish and seafood (24.5%), wild meat (15.6%), mushrooms (12.6%), insects (0%) and algae (0%). For the plant group, the nutritional information is predominantly available at the species level, with 62.4% of foods having species-specific data.

Out of the 36,058 recipes sourced in total from the Household Budget Survey (POF) by the Brazilian Institute of Geography and Statistics (IBGE), cookbooks, and websites, only 72 preparations reported in the POF included ingredients from our list. The remaining recipes were sourced from cookbooks and culinary websites. The distribution of recipes by food group was as follows: plants (60.3%), fish and seafood (20.5%), mushrooms (18.6%), and wild meat (0.6%). Recipes for algae and insects were scarce, with 10 recipes for algae and two for insects. Manihot esculenta Crantz (cassava) emerged as the species with the highest number of associated recipes—6,875, or nearly one-fifth of the total culinary references in the entire database (Supplementary Information 3).

Factors that predict the prioritization in nutritional studies

In both food composition (R²: 0.677) and consumption studies (R²: 0.782), recipe number and species occurrence across different states were the most influential features predicting prioritization (Fig. 2).

Fig. 2.

Priority factors for including species in studies of (A) food composition and (B) food consumption. The beeswarm plot uses SHAP (SHapley Additive exPlanations) values, where positive SHAP values (right) indicate that the feature is contributing to higher prioritization, and negative values (left) favours lower prioritization. Each point represents an observation in the dataset. The color gradient indicates the value of the feature for each observation, with blue points representing lower feature values and red points representing higher feature values. The SHAP values show the impact of each feature on the predicted prioritization score by the model.

Regarding the variables least associated with prioritization (i.e., the other 14 features mentioned in Fig. 2), both in the context of composition and consumption studies, those related to species conservation status, as well as food categories such as insects and algae, exhibited the lowest influence. The variable "Labeler from nutrition science" highlights the differences in prioritization between specialists from nutritional and environmental sciences. Differences in feature-specific prioritization between specialist groups are summarized in Fig. 2.

Both specialist groups agreed on three key variables that should drive decision-making: a high number of recipes and a wide geographical distribution across Brazilian states (Fig. 3). While there is a consensus that recipe number and species occurrence are the most important variables for prioritization, specialists differ in how they prioritize these factors. Environmental science specialists tend to prioritize species with a broader geographic distribution throughout the country, viewing this as the most important factor. In contrast, nutritionists place more emphasis on species with documented culinary uses, as evidenced by their presence in recipes.

Fig. 3.

Comparison of feature prioritization between nutritional and environmental specialists.

While our study identifies key variables and their relative importance for prioritizing species, it does not result in a simple mathematical formula for prioritization. It is because our analysis utilizes a LightGBM model, a complex machine learning algorithm that combines multiple decision trees to make predictions. Extracting a single formula from this intricate structure is impractical and would not accurately represent the model’s decision-making process. To ensure replicability, we provide access to our trained model and data on GitHub. Researchers can apply this model to their datasets, using it to predict species prioritization based on the identified key variables. This approach offers a more nuanced and accurate assessment than a simplified formula could provide.

Discussion

This study addresses the global challenge of prioritizing neglected food species by using Brazil as a case study to develop a framework that is broadly applicable. Our research identified 369 species, primarily plants and wild animals, many of which lack comprehensive nutritional data. The prioritization process revealed that culinary applicability and distribution are the most influential criteria for selecting species. Ecological or conservation aspects were not key factors either, indicating a potential disconnect between nutritional research priorities and broader biodiversity conservation goals. Disciplinary perspectives also played a role, with environmental scientists prioritizing species with wider distribution. These findings underscore that species prioritization for composition and consumption studies is primarily driven by immediate practical utility and researchers’ expertise, potentially overlooking crucial ecological considerations. These insights emphasise that nutritional studies require a more integrated strategy that balances immediate research needs with long-term ecological sustainability.

The prevalence of plants in our inventory reflects their historical dietary importance to early populations in Brazil, revealing a rich and diverse history of plant use that underscores their long-standing role in people’s diets. Archaeobotanical evidence from the Cerrado region suggests that indigenous people utilized a variety of wild plants and later incorporated domesticated species, such as maize, manioc, and beans, reflecting a transition from hunter-gatherer to more settled agricultural practices¹³. This historical reliance on plants is further supported by a systematic review of the use of wild food plants by Brazilian communities, which found that early populations extensively utilized wild plants, with fruits, leaves, and seeds accounting for 85.8% of the plant parts consumed¹⁴. The accessibility of plants likely contributes to their prominence in diets, as opposed to algae, insects, and wild animals. A nationwide study of Brazilian dietary habits revealed that unconventional food plants (UFPs) are consumed more frequently than edible mushrooms and wild meat. Notably, edible algae and insects were excluded from the analysis due to insufficient or non-existent consumption data, highlighting significant gaps in the utilization of these potential food sources².

Building on this understanding of plant prevalence, our study estimates that developing recipes and increasing the availability of nutritional composition data can enhance scientists’ interest in neglected species from local biodiversity, thereby contributing to their popularization and the development of scientific knowledge. The promotion of edible species such as algae, mushrooms, and insects could enhance dietary diversity within primarily plant-based sustainable diets⁵. The benefits of integrating food biodiversity beyond plants into sustainable diets include increased resilience of the food system, better alignment of food cultures with environmental contexts, income generation for local communities, and the revival of cultural knowledge^15,16. However, despite their consumption in many traditional communities, cultural barriers—such as fear of poisoning, familiarity, and food neophobia—continue to limit the acceptance of these food groups by the broader population^17,18.

A key example of cultural barriers to sustainable diets in Brazil is the consumption of insects. This practice remains marginal despite the country harboring the highest diversity of edible insect species in South America. A recent study identified 141 edible insect species in Brazil, with 140 having documented records of consumption; however, many of these records are anecdotal or date back to the colonial period¹⁹. In urban areas, insect consumption is typically driven by curiosity for the “exotic” or by enduring cultural traditions, such as the consumption of ants from the Atta genus, known locally as “tanajura” or “içá”. Some edible species, however, are not consumed in Brazil despite being eaten in other parts of the world, highlighting the potential for expanding this practice. Significant cultural and regulatory challenges impede the broader acceptance of edible insects in Brazil. Insects are often perceived as exotic and evoke neophobia and disgust among Brazilians, particularly among older women with less formal education, while younger, more educated men are somewhat more receptive^20,21. Additionally, the absence of specific regulations on insect consumption raises concerns about food safety and complicates their integration into mainstream diets²². These cultural and institutional barriers present significant obstacles to leveraging insects as a sustainable food source, beyond their occasional use by Indigenous peoples and local communities.

Wild mushrooms, like insects, are largely overlooked in the Brazilian diet despite their nutritional and medicinal value. A recent study identified 409 species of wild edible mushrooms (WEM) in Brazil. However, only 86 have robust records based on molecular data or Brazilian nomenclatural types, highlighting the need for further studies to confirm their identity and occurrence²³. Of these species, 93 are consumed in Brazil, 52 have uncertain or incomplete evidence of consumption, and 264 are not consumed, presenting a potential new food resource. While interest in mushroom foraging is growing, proper training in identifying edible species is crucial to prevent misidentifications and potential poisoning.

Fish and seafood face unique challenges due to inconsistent identification and labeling practices. Many species are sold under generic names, such as "pescada," which can refer to three distinct species in our inventory, obscuring biodiversity and sometimes resulting in the unintentional consumption of threatened species²⁴. This ambiguity complicates assessments of consumption patterns and poses conservation concerns. Furthermore, the nutrient quality of fish and seafood depends heavily on accurate species identification²⁵. Addressing these challenges will require precise labeling practices and comprehensive studies to promote the sustainable use of aquatic food resources.

The reliance on recipe numbers as a primary indicator for species inclusion in nutritional studies demonstrates the need for a more nuanced understanding of cultural food practices, with recipes serving as a valuable source of information. In this regard, our study reveals a significant discrepancy in the documentation of neglected and underutilized foods between national dietary surveys and popular culinary resources. The Brazilian Household Budget Survey reports a mere 0.2% presence of these foods, whereas popular websites feature over 36,000 recipes. This stark contrast highlights the limitations of traditional dietary assessment methods, such as 24-h recalls and food frequency questionnaires, which often underestimate the consumption of episodically consumed foods, including wild, forest, and underutilized species²⁶. This finding emphasizes the urgent need for more comprehensive methods to collect, analyze, and report dietary intake data that accurately reflect the contribution of neglected foods to diets. Given the complexity of developing national surveys focused on food biodiversity in a country as diverse as Brazil, a strategic approach could be to initially focus on communities that rely on traditional food systems, which are recognized for their rich food diversity²⁷. This targeted approach would provide valuable insights into consumption patterns, allowing for the refinement of evaluation tools and the initiation of food composition analyses for these resources. By addressing these gaps, we can break the cycle in which these foods are overlooked due to assumptions about their lack of consumption and excluded from studies due to insufficient compositional data. Beginning with these communities could yield crucial insights and lay the foundation for broader national efforts to integrate food biodiversity into dietary assessments.

This study reveals a generally aligned approach to species prioritization for food research between nutritionists and environmental scientists, with both groups recognizing the importance of recipe prevalence and data availability. However, nuances exist in their perspectives. Nutritionists tend to prioritize species with more recipes available, while environmental scientists may consider broader geographic distribution as an indicator of ecological significance.

Despite these complementary viewpoints, a concerning gap emerges in the lack of emphasis on conservation factors. Importantly, this pattern was not associated with educational background, as it was observed across both areas of expertise. Three main factors may explain this trend. First, this lack of emphasis may stem from the limited availability of robust conservation data; our study found that nearly half of the assessed species lacked information regarding their conservation status. Second, ethical considerations regarding food security may influence expert recommendations. In resource-limited settings, professionals may reasonably avoid discouraging consumption of species that constitute critical nutritional resources for vulnerable populations, even when conservation concerns exist^28,29 (see e.g., ref). Within the framework of our research, the availability of recipe and distribution patterns serves as an indicator of accessibility to these species. Considering Brazil’s considerable prevalence of food insecurity, these factors may play a significant role in explaining experts’ prioritization patterns, as professionals likely weight immediate nutritional needs against long-term conservation objectives when making recommendations about food species utilization. Third, the well-documented implementation gap in conservation science may contribute to this pattern³⁰. Conservation assessments frequently fail to translate into actionable implementation strategies, creating challenges even for experts with strong environmental science backgrounds when attempting to balance conservation priorities with immediate socioeconomic realities.

While practical considerations are important, neglecting the ecological and conservation status of food species risks undermining the long-term sustainability of food systems³¹. Studies should particularly emphasize the interconnected roles of biodiversity in supporting both nutritional health and ecological sustainability. For instance, biodiversity loss in fisheries in the Peruvian Amazon has reduced the availability of key micronutrients, such as zinc and iron, underscoring the importance of conserving diverse species for both dietary stability and ecosystem health³². Conversely, promoting the consumption of species at risk can harm their populations and ecosystems if not properly managed. To ensure that future decisions about species prioritization are informed by a comprehensive understanding of their ecological impact, it is crucial to prioritize the collection and integration of conservation status data, such as extinction risk assessments, into food biodiversity research.

One of the core strengths of this research lies in the explainable AI approach (LightGBM and SHAP analysis), which enabled us to quantify and transparently explain the drivers influencing neglected species prioritization. By identifying the most influential variables in expert decision-making, our analytical framework provides valuable insights into cross-disciplinary prioritization processes. It offers a replicable tool that researchers, funding agencies, NGOs, and government entities can adopt. Beyond the specific application demonstrated here, specialists working with food biodiversity can leverage AI for additional uses. For instance, data imputation is a powerful solution to address missing data, particularly crucial for species with limited records or when ethical constraints impede primary data collection, a significant challenge often faced with species of conservation concern³³.

Among the limitations of this study, we highlight three main points. First, the comprehensive lists of neglected food species were compiled by a core expert panel, which, despite its multidisciplinary nature and national scope, represents a limited group of specialists. This initial reliance on a defined set of experts might introduce a specific perspective, potentially overlooking species or nuances that could be identified by a broader, more diverse pool of contributors. Second, although extensive, the inventory may not be exhaustive and could benefit from supplementation with specialized studies that focus on specific regions or food groups. For example, El Bizri et al.³⁴ assessed the impact of illegal sport hunting on Brazil’s wildlife, documenting the hunting of 157 native species, including 19 endangered ones. Their study revealed significant wildlife depletion, particularly in the Atlantic Forest and Caatinga biomes, with a shift from large mammals to smaller birds as hunting pressure increased. A nutrition study focused on wild meat could leverage our inventory by providing essential data on species, particularly for those that are underutilized and threatened. Each food group, such as wild meat or seafood, may require additional data from specialized studies to capture its complexity fully. This supplementary information would enhance the overall value of our inventory for biodiversity and food consumption research. Third, the prioritization analysis, though informative, relies on expert opinions that may not fully encapsulate the multifaceted complexity of food systems.

We recommend several strategies to promote the consumption of underutilized foods, with practical and policy implications. First, encouraging investment in interdisciplinary and collaborative studies of the nutritional composition of local biodiversity foods is essential to increase their visibility and validate their nutritional value. Second, expanding dietary assessments among communities with limited market integration—who likely consume more diverse diets—could provide valuable insights into traditional dietary practices and support the preservation of food biodiversity. Third, fostering the development of an integrated national database on the culinary uses and nutritional composition of local foods, while strengthening existing initiatives, would be instrumental in making this knowledge more accessible to researchers, policymakers, and the general public. Finally, promoting the inclusion of biodiverse foods in public programs and training professionals working with human nutrition can further support their adoption, aligning dietary diversity with broader public health and environmental sustainability goals. These actions can help bridge the gap between biodiversity conservation and dietary practices, building a more resilient and culturally adaptive food system. Finally, integrating biodiverse foods into public programs and training nutrition professionals are crucial steps toward aligning dietary diversity with public health and environmental sustainability goals. However, it is essential to consider the conservation status of each species and implement measures to prevent ecological impacts. These measures include promoting certified agroecological and agroforestry production, as well as educating consumers on the importance of purchasing from sustainable sources.

This study provides a crucial foundation for understanding and leveraging Brazil’s rich food biodiversity to improve nutrition and sustainable food systems. By compiling a comprehensive inventory of neglected food species and analyzing expert prioritization factors, we have highlighted key areas for future research and policy interventions. Our findings reveal a significant bias towards species with readily available culinary applications, often overlooking the ecological and cultural significance of underutilized food sources. This data underscores the urgent need to shift from a purely nutritional perspective to a more holistic approach that integrates ecological, cultural, and social dimensions of food biodiversity. Prioritizing research on underrepresented food groups, such as algae, insects, and wild mushrooms, particularly those with limited conservation data, is crucial.

Furthermore, integrating traditional knowledge from indigenous peoples and local communities, who possess invaluable insights into the sustainable use of these resources, is essential for developing culturally relevant and ecologically sound food systems. Ultimately, bridging the gap between biodiversity conservation and dietary practices requires a multifaceted approach. It includes investing in interdisciplinary research, expanding dietary assessments to capture traditional foodways, developing comprehensive food databases, and promoting the inclusion of biodiverse foods in public programs. These efforts bridge the gap between biodiversity conservation and dietary practices, fostering a more resilient, culturally adaptive, and ecologically responsible food system.

Methods

Study area

Brazil is the largest tropical country in the world and boasts one of the world’s richest biodiversity, with an estimated 1.8 million species representing 9.5% of the global total³⁵. However, this vast wealth of biodiversity remains untapped and understudied, particularly in terms of its potential contributions to food systems and nutrition. For example, a comprehensive study revealed that only 1% of Brazilians incorporate neglected and underutilized species into their diets, underscoring the enormous unexploited potential of Brazil’s food biodiversity². Furthermore, research focusing on the Brazilian semi-arid, locally known as Caatinga, found that only 20% of its neglected plant species were represented in the country’s primary Food Composition Tables (FTC)³⁶. Additionally, key species from various Brazilian biomes³⁷ are absent from the FCT most widely used by nutrition professionals in the country³⁸, such is the case of baru (Dipteryx alata Vogel) from the Cerrado, carandá (Copernicia alba Morong) from the Pantanal, butiá (Butia odorata (Barb.Rodr.) Noblick) from the Pampa, araçá (Psidium cattleianum Sabine) from the Atlantic Forest, and camu-camu (Myrciaria dubia (Kunth) McVaugh) from the Amazon, which is the richest known plant source of vitamin C globally³⁹. Furthermore, no species of wild mushrooms, insects, or algae are represented in this FCT. These examples highlight not only the limited utilization of Brazil’s diverse food resources but also the significant gaps in our understanding of their nutritional composition, emphasizing the need for more comprehensive research and promotion of these underutilized species in Brazilian diets and food systems⁴⁰.

Creation of lists of neglected foods

A panel of experts was assembled based on their expertise in nutrition and environmental sciences and experience with the following food categories: algae, terrestrial wild vertebrate animals, fish and seafood, insects, mushrooms, and plants. Figure 4 outlines the entire research process, detailing the roles of all actors and steps involved.

Fig. 4.

Study workflow and expert involvement.

A panel of experts was assembled based on their expertise in nutrition and environmental sciences and experience with the following food categories: algae, terrestrial wild vertebrate animals, fish and seafood, insects, mushrooms, and plants. Figure 4 outlines the entire research process, detailing the roles of all actors and steps involved.

The core expert panel, comprising a collaborative effort of 14 individuals, including 13 specialists in various food species groups and one expert in data science, compiled the comprehensive lists of neglected food species across six categories. These ‘core experts’ were invited to participate by the lead researchers (MCMJ and UPA), based on their documented expertise in at least one of the six targeted food species groups. For instance, several panelists had recently published up-to-date systematic reviews or comprehensive lists of neglected food species^23,41.

The panel’s most critical task was to ensure a comprehensive national perspective on these foods. To achieve this, we a systematic approaches, including comprehensive systematic reviews, long-term datasets spanning almost three decades of food consumption across different Brazilian regions, official national data (e.g., from IBAMA), and other official documents. Some categories proved more challenging than others in this regard. For instance, plants encompass a broader and more diverse range of uses compared to other groups, making it more difficult to define “neglected” without introducing critical biases. In such cases, the use of established official lists provided a more objective approach. A prime example is the "Portaria Interministerial MAPA/MMA n. 10/2021," which recognizes native species of food value as sociobiodiversity products in Brazil, serving as a reliable baseline for neglected species classification.

All methodological decisions and detailed procedures for list compilation are documented in Supplementary Material 4. The core expert team has been integral to the research since its inception. Additionally, at least one-third of these experts had previous direct experience working with traditional peoples, particularly indigenous communities, riverine populations, and artisanal fishers. Importantly, in four out of the six food species groups, the team included experts from more than one region of the country.

Selection of features and source of information

The core experts, alongside the lead authors, contributed to the selection of features used to characterize the neglected species. Since no predefined set of features existed, six were identified through a non-systematic literature review (see Box 1) and group discussions. The selected features were: regional occurrence in Brazil (number of Brazilian states with confirmed occurrence), conservation status with nine threat categories (The IUCN Red List of Threatened Species), origin (wild or cultivated), commercial cultivation status (Yes/No), availability of nutritional data (No; Yes, at the species level; Yes, at the genus level; Yes, by common name), and the number of recipes incorporating the ingredient. We implemented a structured approach to collect recipe information for each neglected food species, drawing on primary sources, including the Household Budget Survey (POF/IBGE), cookbooks that focus on neglected species and traditional Brazilian cuisine, and online sources, specifically the three most visited Brazilian recipe websites. Detailed sources for each type of information are provided in Supplementary Information 5. Data collection to populate the database with these features occurred between April and October 2023.

Prioritization and confidence rating process

Our project progressed to a species prioritization and confidence rating phase, which involved a separate group of researchers, referred to as 'ad hoc evaluators.' These evaluators were specialists in either Nutritional or Environmental Sciences and had experience working with at least one of the neglected species categories covered in this study. To streamline recruitment, we developed a standardized form, which was distributed through the professional networks of the research team. All qualified specialists who expressed interest were included in the study. In recognition of their contributions, the evaluators were offered the opportunity to write and review the paper and participate as co-authors in the resulting publication.

In total, 18 ad hoc experts participated, comprising Nutritional Sciences (72.2%) and Environmental Sciences (27.8%), with a balanced distribution of specializations that reflected diverse backgrounds and areas of expertise. In Nutritional Sciences, the experts are evenly split between food composition and food consumption (38.5.8% each), with 23% experts in both fields. Among the food composition experts, many had experience across multiple food categories, with a significant proportion (83%) having extensive experience in plant-based foods, alongside expertise in other categories such as fish, fungi, algae, and insects. In Environmental Sciences, 60.0% of the experts specialize in ethnobiology, while 40.0% combine expertise in both ethnobiology and ecology. These specialists bring diverse knowledge of plants, fish, and terrestrial wildlife. All experts declared no conflicts of interest. In our analysis, the variable "Labeler from nutrition science" highlights the differences in prioritization between specialists from Nutritional and Environmental sciences.

Before beginning the evaluation, ad hoc experts participated in a 30-min orientation session to familiarize themselves with the task. They were also given a detailed dictionary of variables, which included comprehensive descriptions of each variable and, for categorical variables, outlined all possible response options. Then, they were provided with a spreadsheet containing 369 species, each described by six characterization variables (see previous section), along with two prioritization scales and two confidence calibration variables. To prevent bias, the species were not identified by their taxonomy; evaluators were only informed of the general category (algae, terrestrial wild vertebrates, fish, insects, mushrooms, or plants), ensuring that prioritization was based solely on species attributes. Using a 1–5 Likert scale (ranging from very low to very high), the evaluators prioritized the species based on two key factors: their importance for nutritional composition studies and their relevance to dietary consumption research. They also rated their confidence in each response on the same 1–5 scale.

Ethical approval for this study was not required under the Brazilian National Health Council Resolution n. 510/2016, article 1, item VII. This resolution exempts research from Ethics Board review when no personal identification data is collected or published and when the information gathered pertains solely to the participants’ expertise rather than personal matters, as is the case with this study.

Data analysis

We examined the traits of the species included in the inventory using descriptive statistics, including metrics such as mean, median, standard deviation, and interquartile range to gain an understanding of the patterns and variability within the data. These data provide insights into various characteristics, including habitat types, geographic distributions, and physical traits of the species.

We then employed a supervised machine learning regression model to investigate how these parameters influenced the order of prioritization for composition and consumption research, and to forecast prioritization scores based on the traits of the species. This method allowed us to capture and analyze the complex relationships between multiple independent variables (the species characteristics) and the dependent variable (the prioritization scores).

Then, we performed tenfold cross-validation⁴² to ensure our model was reliable and broadly applicable. This method divides the dataset into ten equally-sized pieces, or "folds." After that, the model is trained ten times, with the training set consisting of the remaining nine folds and the validation set consisting of a different fold each time. By reducing variation and bias in the model’s performance estimation, this procedure offers a more accurate estimation of the model’s potential performance on unseen data.

Additionally, we selected the Light Gradient Boosting Machine (LightGBM) algorithm⁴³ as our modeling framework for the regression analysis. Using a gradient boosting architecture, LightGBM creates decision trees sequentially to correct the mistakes of previous ones. It is widely used due to its effectiveness and speed. We have selected LightGBM after assessing several commonly available architectures and determining that it had the lowest Root Mean Squared Error (RMSE) in our data.

After training our final LightGBM model, we employed SHAP (SHapley Additive exPlanations)⁴⁴ to analyze the model’s outputs and understand how each feature influenced the predictions. Employing ideas from cooperative game theory, SHAP values calculate the contribution of each feature to explain a particular forecast. SHAP increases the transparency of the model’s decision-making process by assigning an importance value to each feature for a specific prediction.

To visualize SHAP data, we employed a variety of plots, such as force, dependence, and summary plots. With the use of these visualizations, we were able to determine not only which characteristics had the greatest overall influence but also how modifications to a certain feature’s value could impact the prioritization score. This level of interpretability was essential for both validating our model and providing insights for future research.

We used several specialized Python libraries in our analysis. To process the datasets, we used Pandas⁴⁵. We performed cross-validation, feature extraction, and preprocessing using scikit-learn⁴⁶. The gradient boosting algorithm originated from the LightGBM package⁴⁷. Using the SHAP library⁴⁸, we computed and displayed SHAP values. We assessed the model using traditional regression metrics, mainly the Root Mean Squared Error (RMSE), and the R-squared (R²) coefficient. These metrics provided numerical assessments of the model’s precision and ability to account for variation in the priority scores. The database and analysis scripts are available at this link https://github.com/eliasjacob/paper_bionut/

Author contributions

MCMJ led the project with responsibilities in Conceptualization, Data curation, Formal analysis, Funding acquisition, Methodology, Project administration, Validation, Visualization, and Writing (original draft and review & editing). UPA provided Conceptualization, Supervision, and Writing (review & editing). EJMN contributed to Conceptualization, Formal analysis, Methodology, Visualization, and Writing (review & editing). DT, DVBO, FOP, GFM, JKMS, LJGS, MPD, NMJ, PFML, JVMS, SCGS and RRVS contributed to Data Curation, Conceptualization and Writing (review & editing). AMC, SMG, and DMLM contributed to Conceptualization and Writing (review & editing). AGB, AFSJ, AAFC, CMMM, CBBC, EBG, HREB, IWLB, JSA, JS, JAT, LZOC, MAO, MEGO, MMN, RRNA, SCVCL, TQM contributed to Validation and Writing (review & editing). All authors have reviewed and approved the final manuscript and agree to be accountable for all aspects of the work.

Funding

This research was supported by the High-Performance Computing Center at UFRN (NPAD/UFRN) and the ‘Conselho Nacional de Desenvolvimento Científico e Tecnológico’ (CNPq) through a research grant to MCMJ (402334/2021–3) and AMC (444588/2023–0). The authors also thank CNPq for the research productivity scholarships awarded to EJMN (302582/2023–1), MCMJ (306755/2021–1), NMJ (314236/2021–0), and PFML (302365/2022–2). PFML acknowledges the funding from the Romanian Ministry of Research, Innovation and Digitalization for project JUST4MPA (760054, within the framework of PNRR-III-C9-2022—I8). AMC, MPD, and NMJ acknowledge the support of the Sao Paulo Research Foundation – FAPESP by, respectively, the following grants: 2022/03091–6, 2017/25754–9, and 2018/15677–0. HREB was supported by USAID agreement with CIFOR-ICRAF. CBBC also acknowledges the support from CNPq (306891/2021–2) and from the ‘Coordenação de Aperfeiçoamento de Pessoal de Nível Superior’ (CAPES) – Finance Code 001. The study received a contribution from the INCT Ethnobiology, Bioprospecting, and Nature Conservation, certified by CNPq, with financial support from the Foundation for Support to Science and Technology of the State of Pernambuco as a grant [Grant Number: APQ-0562–2.01/17] given to UPA.

Data availability

The database and analysis scripts are available at this link https://github.com/eliasjacob/paper_bionut.

Competing interests

The authors declare no competing interests.