Artificial Intelligence in Functional Food Ingredient Discovery and Characterisation: A Focus on Bioactive Plant and Food Peptides.

Scientific research consistently demonstrates that diseases may be delayed, treated, or even prevented and, thereby, health may be maintained with health-promoting functional food ingredients (FFIs). Consumers are increasingly demanding sound information about food, nutrition, nutrients, and their associated health benefits. Consequently, a nutrition industry is being formed around natural foods and FFIs, the economic growth of which is increasingly driven by consumer decisions. Information technology, in particular artificial intelligence (AI), is primed to vastly expand the pool of characterised and annotated FFIs available to consumers, by systematically discovering and characterising natural, efficacious, and safe bioactive ingredients (bioactives) that address specific health needs. However, FFI-producing companies are lagging in adopting AI technology for their ingredient development pipelines for several reasons, resulting in a lack of efficient means for large-scale and high-throughput molecular and functional ingredient characterisation. The arrival of the AI-led technological revolution allows for the comprehensive characterisation and understanding of the universe of FFI molecules, enabling the mining of the food and natural product space in an unprecedented manner. In turn, this expansion of bioactives dramatically increases the repertoire of FFIs available to the consumer, ultimately resulting in bioactives being specifically developed to target unmet health needs.

An Artificial-Intelligence-Discovered Functional Ingredient, NRT_N0G5IJ, Derived from Pisum sativum, Decreases HbA1c in a Prediabetic Population.

The prevalence of prediabetes is rapidly increasing, and this can lead to an increased risk for individuals to develop type 2 diabetes and associated diseases. Therefore, it is necessary to develop nutritional strategies to maintain healthy glucose levels and prevent glucose metabolism dysregulation in the general population. Functional ingredients offer great potential for the prevention of various health conditions, including blood glucose regulation, in a cost-effective manner. Using an artificial intelligence (AI) approach, a functional ingredient, NRT_N0G5IJ, was predicted and produced from Pisum sativum (pea) protein by hydrolysis and then validated. Treatment of human skeletal muscle cells with NRT_N0G5IJ significantly increased glucose uptake, indicating efficacy of this ingredient in vitro. When db/db diabetic mice were treated with NRT_N0G5IJ, we observed a significant reduction in glycated haemoglobin (HbA1c) levels and a concomitant benefit on fasting glucose. A pilot double-blinded, placebo controlled human trial in a population of healthy individuals with elevated HbA1c (5.6% to 6.4%) showed that HbA1c percentage was significantly reduced when NRT_N0G5IJ was supplemented in the diet over a 12-week period. Here, we provide evidence of an AI approach to discovery and demonstrate that a functional ingredient identified using this technology could be used as a supplement to maintain healthy glucose regulation.

Ageing transcriptome meta-analysis reveals similarities and differences between key mammalian tissues.

By combining transcriptomic data with other data sources, inferences can be made about functional changes during ageing. Thus, we conducted a meta-analysis on 127 publicly available microarray and RNA-Seq datasets from mice, rats and humans, identifying a transcriptomic signature of ageing across species and tissues. Analyses on subsets of these datasets produced transcriptomic signatures of ageing for brain, heart and muscle. We then applied enrichment analysis and machine learning to functionally describe these signatures, revealing overexpression of immune and stress response genes and underexpression of metabolic and developmental genes. Further analyses revealed little overlap between genes differentially expressed with age in different tissues, despite ageing differentially expressed genes typically being widely expressed across tissues. Additionally we show that the ageing gene expression signatures (particularly the overexpressed signatures) of the whole meta-analysis, brain and muscle tend to include genes that are central in protein-protein interaction networks. We also show that genes underexpressed with age in the brain are highly central in a co-expression network, suggesting that underexpression of these genes may have broad phenotypic consequences. In sum, we show numerous functional similarities between the ageing transcriptomes of these important tissues, along with unique network properties of genes differentially expressed with age in both a protein-protein interaction and co-expression networks.

A scan for genes associated with cancer mortality and longevity in pedigree dog breeds.

Selective breeding of the domestic dog (Canis lupus familiaris) rigidly retains desirable features, and could inadvertently fix disease-causing variants within a breed. We combine phenotypic data from > 72,000 dogs with a large genotypic dataset to search for genes associated with cancer mortality and longevity in pedigree dog breeds. We validated previous findings that breeds with higher average body weight have higher cancer mortality rates and lower life expectancy. We identified a significant positive correlation between life span and cancer mortality residuals corrected for body weight, implying that long-lived breeds die more frequently from cancer compared to short-lived breeds. We replicated a number of known genetic associations with body weight (IGF1, GHR, CD36, SMAD2 and IGF2BP2). Subsequently, we identified five genetic variants in known cancer-related genes (located within SIPA1, ADCY7 and ARNT2) that could be associated with cancer mortality residuals corrected for confounding factors. One putative genetic variant was marginally significantly associated with longevity residuals that had been corrected for the effects of body weight; this genetic variant is located within PRDX1, a peroxiredoxin that belongs to an emerging class of pro-longevity associated genes. This research should be considered as an exploratory analysis to uncover associations between genes and longevity/cancer mortality.

From humans to hydra: patterns of cancer across the tree of life.

Cancer is a disease of multicellularity; it originates when cells become dysregulated due to mutations and grow out of control, invading other tissues and provoking discomfort, disability, and eventually death. Human life expectancy has greatly increased in the last two centuries, and consequently so has the incidence of cancer. However, how cancer patterns in humans compare to those of other species remains largely unknown. In this review, we search for clues about cancer and its evolutionary underpinnings across the tree of life. We discuss data from a wide range of species, drawing comparisons with humans when adequate, and interpret our findings from an evolutionary perspective. We conclude that certain cancers are uniquely common in humans, such as lung, prostate, and testicular cancer; while others are common across many species. Lymphomas appear in almost every animal analysed, including in young animals, which may be related to pathogens imposing selection on the immune system. Cancers unique to humans may be due to our modern environment or may be evolutionary accidents: random events in the evolution of our species. Finally, we find that cancer-resistant animals such as whales and mole-rats have evolved cellular mechanisms that help them avoid neoplasia, and we argue that there are multiple natural routes to cancer resistance.

A new approach for interpreting Random Forest models and its application to the biology of ageing.

Motivation: This work uses the Random Forest (RF) classification algorithm to predict if a gene is over-expressed, under-expressed or has no change in expression with age in the brain. RFs have high predictive power, and RF models can be interpreted using a feature (variable) importance measure. However, current feature importance measures evaluate a feature as a whole (all feature values). We show that, for a popular type of biological data (Gene Ontology-based), usually only one value of a feature is particularly important for classification and the interpretation of the RF model. Hence, we propose a new algorithm for identifying the most important and most informative feature values in an RF model. Results: The new feature importance measure identified highly relevant Gene Ontology terms for the aforementioned gene classification task, producing a feature ranking that is much more informative to biologists than an alternative, state-of-the-art feature importance measure. Availability and implementation: The dataset and source codes used in this paper are available as 'Supplementary Material' and the description of the data can be found at: https://fabiofabris.github.io/bioinfo2018/web/. Supplementary information: Supplementary data are available at Bioinformatics online.

Identification of polymorphisms in cancer patients that differentially affect survival with age.

The World Health Organization predicts that the proportion of the world's population over 60 will almost double from 12% to 22% between 2015 and 2050. Ageing is the biggest risk factor for cancer, which is a leading cause of deaths worldwide. Unfortunately, research describing how genetic variants affect cancer progression commonly neglects to account for the ageing process. Herein is the first systematic analysis that combines a large longitudinal data set with a targeted candidate gene approach to examine the effect of genetic variation on survival as a function of age in cancer patients. Survival was significantly decreased in individuals with heterozygote or rare homozygote (i.e. variant) genotypes compared to those with a common homozygote genotype (i.e. wild type) for two single nucleotide polymorphisms (rs11574358 and rs4147918), one gene (SIRT3) and one pathway (FoxO signalling) in an age-dependent manner. All identified genes and pathways have previously been associated with ageing and cancer. These observations demonstrate that there are ageing-related genetic elements that differentially affect mortality in cancer patients in an age-dependent manner. Understanding the genetic determinants affecting prognosis differently with age will be invaluable to develop age-specific prognostic biomarkers and personalized therapies that may improve clinical outcomes for older individuals.

Comparison of sodium content of meals served by independent takeaways using standard versus reduced holed salt shakers: cross-sectional study.

BACKGROUND: Takeaway food has a relatively poor nutritional profile. Providing takeaway outlets with reduced-holed salt shakers is one method thought to reduce salt use in takeaways, but effects have not been formally tested. We aimed to determine if there was a difference in sodium content of standard fish and chip meals served by Fish & Chip Shops that use standard (17 holes) versus reduced-holed (5 holes) salt shakers, taking advantage of natural variations in salt shakers used. METHODS: We conducted a cross-sectional study of all Fish & Chip Shops in two local government areas (n = 65), where servers added salt to meals as standard practice, and salt shaker used could be identified (n = 61). Standard fish and chip meals were purchased from each shop by incognito researchers and the purchase price and type of salt shaker used noted. Sodium content of full meals and their component parts (fish, chips, and fish batter) was determined using flame photometry. Differences in absolute and relative sodium content of meals and component parts between shops using reduced-holed versus standard salt-shakers were compared using linear regression before and after adjustment for purchase price and area. RESULTS: Reduced-holed salt shakers were used in 29 of 61 (47.5 %) included shops. There was no difference in absolute sodium content of meals purchased from shops using standard versus reduced-holed shakers (mean = 1147 mg [equivalent to 2.9 g salt]; SD = 424 mg; p > 0.05). Relative sodium content was significantly lower in meals from shops using reduced-holed (mean = 142.5 mg/100 g [equivalent to 0.4 g salt/100 g]; SD = 39.0 mg/100 g) versus standard shakers (mean = 182.0 mg/100 g; [equivalent to 0.5 g salt/100 g]; SD = 68.3 mg/100 g; p = 0.008). This was driven by differences in the sodium content of chips and was extinguished by adjustment for purchase price and area. Price was inversely associated with relative sodium content (p < 0.05). CONCLUSIONS: Using reduced-holed salt shakers in Fish & Chip Shops is associated with lower relative sodium content of fish and chip meals. This is driven by differences in sodium content of chips, making our results relevant to the wide range of takeaways serving chips. Shops serving higher priced meals, which may reflect a more affluent customer base, may be more likely to use reduced-holed shakers.

Has gene duplication impacted the evolution of Eutherian longevity?

One of the greatest unresolved questions in aging biology is determining the genetic basis of interspecies longevity variation. Gene duplication is often the key to understanding the origin and evolution of important Eutherian phenotypes. We systematically identified longevity-associated genes in model organisms that duplicated throughout Eutherian evolution. Longevity-associated gene families have a marginally significantly higher rate of duplication compared to non-longevity-associated gene families. Anti-longevity-associated gene families have significantly increased rate of duplication compared to pro-longevity gene families and are enriched in neurodegenerative disease categories. Conversely, duplicated pro-longevity-associated gene families are enriched in cell cycle genes. There is a cluster of longevity-associated gene families that expanded solely in long-lived species that is significantly enriched in pathways relating to 3-UTR-mediated translational regulation, metabolism of proteins and gene expression, pathways that have the potential to affect longevity. The identification of a gene cluster that duplicated solely in long-lived species involved in such fundamental processes provides a promising avenue for further exploration of Eutherian longevity evolution.

Population genomics reveal recent speciation and rapid evolutionary adaptation in polar bears.

Polar bears are uniquely adapted to life in the High Arctic and have undergone drastic physiological changes in response to Arctic climates and a hyper-lipid diet of primarily marine mammal prey. We analyzed 89 complete genomes of polar bear and brown bear using population genomic modeling and show that the species diverged only 479-343 thousand years BP. We find that genes on the polar bear lineage have been under stronger positive selection than in brown bears; nine of the top 16 genes under strong positive selection are associated with cardiomyopathy and vascular disease, implying important reorganization of the cardiovascular system. One of the genes showing the strongest evidence of selection, APOB, encodes the primary lipoprotein component of low-density lipoprotein (LDL); functional mutations in APOB may explain how polar bears are able to cope with life-long elevated LDL levels that are associated with high risk of heart disease in humans.

Translational selection frequently overcomes genetic drift in shaping synonymous codon usage patterns in vertebrates.

Synonymous codon usage patterns are shaped by a balance between mutation, drift, and natural selection. To date, detection of translational selection in vertebrates has proven to be a challenging task, obscured by small long-term effective population sizes in larger animals and the existence of isochores in some species. The consensus is that, in such species, natural selection is either completely ineffective at overcoming mutational pressures and genetic drift or perhaps is effective but so weak that it is not detectable. The aim of this research is to understand the interplay between mutation, selection, and genetic drift in vertebrates. We observe that although variation in mutational bias is undoubtedly the dominant force influencing codon usage, translational selection acts as a weak additional factor influencing synonymous codon usage. These observations indicate that translational selection is a widespread phenomenon in vertebrates and is not limited to a few species.

Increased genome sampling reveals a dynamic relationship between gene duplicability and the structure of the primate protein-protein interaction network.

Although gene duplications occur at a higher rate, only a small fraction of these are retained. The position of a gene's encoded product in the protein-protein interaction network has recently emerged as a determining factor of gene duplicability. However, the direction of the relationship between network centrality and duplicability is not universal: In Escherichia coli, yeast, fly, and worm, duplicated genes more often act at the periphery of the network, whereas in humans, such genes tend to occupy the most central positions. Herein, we have inferred duplication events that took place in the different branches of the primate phylogeny. In agreement with previous observations, we found that duplications generally affected the most central network genes, which is presumably the process that has most influenced the trend in humans. However, the opposite trend--that is, duplication being more common in genes whose encoded products are peripheral in the network--is observed for three recent branches, including, quite counterintuitively, the external branch leading to humans. This indicates a shift in the relationship between centrality and duplicability during primate evolution. Furthermore, we found that genes encoding interacting proteins exhibit phylogenetic tree topologies that are more similar than expected for random pairs and that genes duplicated in a given branch of the phylogeny tend to interact with those that duplicated in the same lineage. These results indicate that duplication of a gene increases the likelihood of duplication of its interacting partners. Our observations indicate that the structure of the primate protein-protein interaction network affects gene duplicability in previously unrecognized ways.