Key Messages

• Cognitive and behavioral traits emerge from complex, interconnected networks of biological molecules and pathways within the brain and body.
• These pathways are dynamic and influenced by both intrinsic factors (e.g., genetic ancestry and hereditary variability) and extrinsic factors (e.g., stress, nutrition, environmental contaminants, and social interactions), resulting in variable cognitive performance across different conditions and ages.
• Disruptions to the biological pathways underlying cognition and behavior cascade to affect performance in ways that may not fully be captured or explained by simple gene-behavior correlations, such as those revealed by genome-wide association studies and genomics.
• Multiomics, which integrates data from other omics techniques (e.g., transcriptomics, epigenomics, proteomics, and metabolomics), looks beyond genes and is being used to elucidate some of the molecular profiles underlying complex traits (i.e., the phenome) that are shaped through interactions between genes and environment. It has found applications for precision medicine and precision agriculture and is particularly well suited to study the molecular and pathway-level determinants of human cognition and behavior.
• Most studies to date have established only correlative, rather than causative, links to human cognitive performance, primarily relying on genomic or other omic associations. Improved methods and research for causal inference will be important for identifying molecular targets to develop biological interventions.
• A more complete understanding of the link between gene expression and cognition and behavior involves untangling multiple levels of complexity and elucidating interactions between multiple molecules and signal pathways within and between brain circuits and regions.
• Future directions include research to understand the relationship between molecular profiles and brain circuits, and the contribution of these interactions within the brain and body to cognitive and behavioral phenotypes.
• Similarly, the integration of complementary techniques and data streams, such as neuroimaging, along with research exploring how confounding intrinsic and external factors interact to influence molecular mechanisms and profiles could enhance current understanding of cognitive performance.

This issue paper focuses on the use of multiomics as a tool to understand and identify molecular and pathway-level determinants of human cognitive and behavioral traits, particularly in relation to modifying human performance. It also addresses several technical, practical, security, and principles-based considerations associated with these analyses and their use in developing interventions.

Modifying Human Performance

Human performance modification is described as either enhancement or diminishment of human physical and cognitive performance through various means (NRC, 2012). Early approaches to modifying human performance included biochemical augmentation (e.g., using biologically active molecules or pharmaceuticals) and physical or behavioral augmentation (e.g., training to enhance specific capabilities of interest to the military) (NRC, 1988). In the mid-1980s, scientific knowledge and technologies did not support many of the approaches suggested for human performance modification. However, as research in this area advanced in the 1990s and early 2000s, new methods and emerging capabilities provided novel opportunities for enhancing human performance. Some approaches used computational technologies, such as augmented reality and artificial intelligence (AI) to improve decision-making and assessments of decisions, and a commercial device that could “interfere with and prevent speech production” (NRC, 2012; Shao et al., 2021). Biological approaches also were leveraged, including the engineering of tissues to enhance repair and regeneration, administering of pharmaceuticals for managing fatigue, and use of brain-computer technologies to detect and direct neural activity (NASEM, 2022).

Biotechnologies, their underlying scientific foundations, and the convergence with engineering and computation with the life sciences, increasingly have been viewed as providing new opportunities for modulating human performance by militaries around the world (Kania, 2021).¹ The production of bio-based materials using engineering biology approaches (e.g., the design-build-test-learn cycle) to produce strong, flexible, and responsive clothing represents only one example application that could be used for protective garments, smart fabrics, artificial muscle, and water-driven actuators (Tadjdeh, 2016; Xu, 2024). The United States and China are both active in research and development of bio-derived or -inspired textiles (Chen et al., 2023; Xu, 2024). More recently, AI and machine learning (ML) have been coupled to brain-computer devices (invasive, non-invasive, and wearable) to enhance users’ hands-free control of external objects and to stimulate neural function (Blumenthal et al., 2021; Shao et al., 2021; NASEM, 2022). Similarly, advances in wearable and ingestible devices, some of which may be made with biocompatible materials, increasingly are being used to sense, measure, and signal various physiological and biochemical changes in the user’s body and/or external environment (Blumenthal et al., 2021; NASEM, 2023b). In addition, recent studies suggest that the gut microbiome influences neural activity and cognition, presenting another possible avenue for modulation with potential clinical applications for treatment of cognitive disorders (Mohajeri et al., 2018; Tooley, 2020; NASEM, 2023a; Castells-Nobau et al., 2024). However, the mechanisms of action are poorly understood.

In 2014, the emergence of more precise genome editing (i.e., clustered regularly interspaced short palindromic repeats, CRISPR) in cells to correct disease-causing mutations elicited questions about its potential use for enhancing traits, such as strength and intelligence, or reducing disease susceptibility (DiEuliis and Giordano, 2017a, 2017b; Blumenthal et al., 2021). Although some phenotypes may be tunable through the modification of one gene, many traits, including those involved in cognitive function, are associated with more than one gene (i.e., polygenic), necessitating polygenic editing. Heritable polygenic

___________________

editing could become feasible in the next several decades to treat polygenic diseases and has been touted as a potential next frontier for genomic medicine (Visscher, 2025).

To enable precise and targeted human performance modification, either through gene editing techniques or other interventions, molecular targets and pathways contributing to the desired trait first must be determined. Attempts to identify specific molecular markers or targets in the human genome that might enable such modifications often have involved genome-wide association studies (GWAS), which have been met with variable success, likely because many traits are influenced by multiple genes and function in dynamic, combinatorial, and interactive biological networks. A limited understanding of how different external stimuli (e.g., chemicals and pollutants, extreme temperature, radiation) affect traits in individuals of differing genetic backgrounds presents significant challenges in pinpointing specific molecular determinants of those traits (Ottoman, 1996). External stimuli have been shown to induce molecular and pathway-level changes that affect cognitive performance (Rochester et al., 2018; Tamiz et al., 2022; Banerjee et al., 2024).

Multiomics is well suited to parse the complexity of biological networks and pathways by integrating information about other types of biomolecular players (e.g., transcripts, proteins, metabolites), their modifications, and interactions in time and space (organelles, cells, tissues, brain regions) and may enhance current understanding about molecular determinants of cognitive and behavioral traits. When identifying molecular determinants, advanced statistical, computational, and bioinformatics approaches recently have been used to analyze and integrate multiomics data to better understand complex traits, or the “phenome” for various diseases, including neurological disorders, cognition, and behavior. In addition to multiomics, several other technologies have emerged to better understand the complexity of networks in the human brain as they relate to cognition and behavior. Recent breakthroughs in neuroimaging have enabled more precise mapping of the human brain and characterization of traits, such as intelligence, based on unique individual brain connectivity (Shen et al., 2017; NASEM, 2022; Samardzija et al., 2024) and has used connectome fingerprinting to identify individuals with remarkable accuracy (Finn, 2015). Additionally, invasive and non-invasive (including wearable) brain-computer technologies have led to an increase in data on electrical signals in the brain in various contexts and situations (NASEM, 2022).

Linking Genes to Cognition: Positive and Negative Case Studies

Although several single gene variants have been linked to significant reductions in cognitive performance, no single gene has been found to have a major impact on enhancing cognitive performance. Instead, it is more common for multiple genes to be associated with modest effects on cognition and behavior across the full spectrum of performance. Understanding the roles of genes and other biological molecules within interconnected pathways in mediating cognition and behavior requires knowledge about how these molecules regulate brain structures and functions across various scales, levels, and timeframes. As with the combinatorial and interactive operation of biological pathways, cognition and behavior emerge from complex interactions between multiple brain circuits and regions. However, if the relationship between

gene expression² in the context of brain circuits can be determined, then its role in mediating behavior can be better understood. Two case studies illustrate instances where a gene could and could not be linked to a specific brain circuit and behavioral outcome.

For instance, working memory, a type of short-term memory important for task execution, has shown great variation in animals. Working memory relies on a particular circuit involving a connection between two parts of the brain, the thalamus and prefrontal cortex, but the genetic basis for the role of the thalamus was not well understood until recently. A study that conducted genetic mapping of these brain areas in 200 genetically diverse mice revealed that one gene, Gpr12, accounts for 17% of the variance in the working memory of mice (Hsiao et al., 2020). This gene encodes an orphan receptor expressed in the thalamus, for which the ligand is currently unknown. This example illustrates a case for which a gene expressed in a particular brain region could be associated with variation in working memory.

A counterexample to the ability to link the expression of specific genes in specific brain regions with behavioral outcomes is the large-scale genetic studies of autism spectrum disorder (ASD) based on the Simons Simplex Collection³ (SSC) (Fischbach and Lord, 2010). As implied by its name, patients with ASD exhibit a wide diversity of phenotypes, and hundreds of genes have been associated with the disorder (Sahin and Sur, 2015). Variants of these genes have been associated with specific behavioral outcomes but are expressed ubiquitously in the brain and other parts of the body, precluding direct linking to any particular brain region (Zhou et al., 2022; Rolland et al., 2023; Ali et al., 2025; Andersen et al., 2025). Many investigators have attempted to attribute the behavioral effects of genes to different brain regions, including the neocortex, hippocampus, and striatum. The mode of interaction between brain regions could be a common thread in the operation of these different brain areas, but accounting for the variety of behavioral outcomes for variants in even a particular gene is considered a grand challenge in genetic behavioral research. Certain circuits may be more vulnerable to the dysfunction of a gene. However, as noted previously, although some single gene variants have been causally associated with reduced cognitive ability, this trend does not necessarily translate to enhanced cognitive ability. Functional outcomes also could depend on the interaction between that gene and the variation offset of other genes with which that gene interacts. The case of ASD illustrates just how challenging understanding the link between genes and cognition can be.

In short, a more complete understanding of the link between gene expression and cognition and behavior involves untangling multiple levels of complexity and elucidating the interaction between multiple signal transduction pathways and between multiple brain circuits and regions.

___________________

Multiomics for Uncovering Molecular Insights on the Phenome

Now more than ever, scientists recognize the interplay between intrinsic and extrinsic factors and the important roles that other biological molecules (e.g., RNA, proteins, lipids, metabolites), their modifications (e.g., DNA and RNA methylation, histone modification, protein phosphorylation), and the dynamic, combinatorial network interactions influencing complex traits. Advancements in omics technologies and methods—including epigenomics (chemical changes to DNA and histone proteins that affect gene expression), transcriptomics and epitranscriptomics (transcripts and different RNA subtypes, and their modifications, respectively), proteomics (proteins and their proteoforms), and metabolomics (metabolites)—have greatly improved the ability to measure and understand biological processes at the systems-level. The combination of two or more of these single -omic methods (including genomics) is referred to as “multiomics.” Progress in statistical, computational, and bioinformatics techniques have enhanced the ability to analyze and integrate multiomics data, helping to uncover insights about biological pathways (e.g., cell signaling) influencing traits such as cognitive function. These collective advancements have transformed the understanding of the molecular factors that contribute to complex traits, referred to as the “phenome.”

During the past 10 years, in-depth phenotyping using multiomics has provided molecular-level insight into physiological processes and pathologies (Halama et al., 2024). Multiomics recently has been applied to investigate molecular changes in humans in response to factors such as viral infection (Mikaeloff et al., 2023), exercise (Contrepois et al., 2020), and even space flight (Garrett-Bakelman and Dershi, 2019). Studies have applied multiomics to find molecular markers of athletic performance (Pitsiladis et al., 2016) and to understand various human diseases, such as diabetes (Halama et al., 2024) and cancer (Ghandi et al., 2019; Granga et al., 2019). Spatiotemporal multiomics can provide further resolution of molecular profiles across timescales and spatial dimensions (Liu et al., 2024) and have been used to study different brain regions (Fangma et al., 2023; Hong et al., 2023).

Although powerful, multiomics can be resource-intensive and time consuming. One group has shown that some multiomic traits may be predicted genetically, representing a cost-effective alternative to large-scale multiomic studies (Xu et al., 2023). They used ML to train genetic information of more than 17,000 molecular traits based on proteomic, transcriptomic, and metabolomic data. Indeed, multiomic studies have enabled targeted identification of molecular determinants and genetic risk factors, which have in turn informed precision medicine to improve patient outcomes (Pammi, 2023; Mukherjee et al., 2024) and precision agriculture to enhance crop production (Zhang, 2022). Because multiomic approaches are well suited for studying complex traits affected by both intrinsic and extrinsic factors, they have increasingly been applied to understand complex cognitive and behavioral traits (Ashbrook, 2018) and neurological diseases and disorders, such as Alzheimer’s disease (Shi, 2023), ASD (Freedman, 2023), and schizophrenia (Campeau, 2021).

Associating Cognitive and Behavioral Traits with Molecular Determinants

Integrating other types of molecular datasets with the findings from GWAS have deepened insights into the biology behind complex neurological phenotypes (Hatcher et al., 2019; Kibinge et al., 2020; Korouglou et al., 2023; Wu et al., 2023; Gupta et al., 2024). To better understand the genetic architecture of five “key” personality traits—neuroticism, extraversion, agreeableness, conscientiousness, and openness—researchers conducted GWAS on the Million Veteran Program cohort and identified a variety of independent genome-wide significant loci associated with these traits (Gupta et al., 2024). Gene-based association testing revealed 254 genes showing significant association with at least one of the five personality traits, if not multiple. They followed this analysis with transcriptome-wide and proteomic-wide association (TWAS and PWAS, respectively) to find certain genes and proteins with altered expression. Enrichment analysis of the biological pathways involved for those altered molecular profiles revealed deeper insight into the intricate biology underlying personality traits.

Amplifying the recognition of trait complexity is a GWAS analysis on insomnia, in which genetic data on 386,533 individuals from the United Kingdom Biobank and genetic data on 944,477 individuals from 23andMe were analyzed (Jansen et al., 2019). The study revealed no less than 202 genomic regions, suggesting the involvement of more than 900 genes. These results emphasize that molecular contributions occur against the backdrop of human genetic variability and that molecular factors extend beyond genetics alone. For these reasons, a strong case can be made for multiomic approaches that include epigenomics, transcriptomics, proteomics, metabolomics, or some mixture, to enhance the understanding of the underlying biology and how to perturb or tune pathways to have the desired cognitive or behavioral impact, whether for preventing or treating disease or for modulating performance. Multiomics studies that examine cellular, tissue, and whole-body responses to external factors are broadly referred to as “exposomics.” These studies have contributed valuable insights into how extrinsic factors can influence biological processes and impact various traits, including cognitive function.

Technical, Practical, Security, and Principles-Based Considerations

One technical challenge with GWAS and other omic-wide association studies is that many of these analyses are based on statistical associations and are correlative, rather than causative. Improved methodology for causal inference could help address this challenge for future studies. However, at present, this challenge presents difficulties in developing interventions that specifically restore diminished function without the risk of off-target or unanticipated effects. Although any intervention (invasive or not) has the potential to cause harm, these side effects are generally well understood and often balanced by the benefits. However, in the case of performance enhancement, balancing adverse effects with potential benefits presents unique security and ethical challenges. Another principles-based challenge is the balance and use of resources toward

research and development for cognitive enhancement of healthy individuals versus treatment of cognitive disorders and diseases.

Another practical challenge of using multiomics as a tool to understand cognition and behavior is the use of cell lines and non-human models and potential confounding factors. For instance, behaviors observed in wild-derived animals are not always observed in laboratory animals, and behavioral traits in laboratory animals, which often are inbred, may be more extreme than in wild animals (Ashbrook et al., 2018). For studies using cell lines, the age of cells affects transcriptome levels and characteristics (Skene et al., 2017). Finally, although these technologies have clear applications for advancing an understanding of biological systems toward advancing health, enhancing human performance across several dimensions raises security and principles-based considerations (DiEuliis and Giordano, 2017a, 2017b).

Building on years of research at the intersection of human genomics and precision medicine, several other issues have been raised that pertain to the use of multiomics to enhance or diminish cognition and behavior. Examples of these issues include: (a) return of results to people involved in research, specifically the responsibility of scientists to share and explain the results of research to those involved (NASEM, 2018, 2023c); (b) sharing of incidental findings, specifically results that go beyond the scope of the study that may indicate a health risk to people participating in research (McGuire et al., 2013); (c) recruitment of specific populations of individuals in studies (Parker, 2022); and (d) future analysis of genomic data collected from research or direct-to-consumer services (Bathe and McGuire, 2009). As these discussions were taking place for precision medicine, scientists were debating the principles-based considerations associated with precise editing of the human germline, including editing of early human embryos, sperm, eggs, and cells that become germ cells (NASEM, 2017). These experts concluded that human germline editing may be permissible under strict oversight if no reasonable alternatives exist, the edits are used only to prevent serious disease or conditions in people who have demonstrated or are strongly predisposed to those diseases, and the edits only correct a disease (or, condition)-causing version of a gene to a version not associated with the disease or condition. This conclusion aligned with a contemporary survey of the general public indicating that most respondents support editing of harmful mutations that threaten an individual’s health either at birth (72% support) or later on in life (60% support), but not for enhancing a particular trait, such as intelligence (19% support).⁴ Although these reports did not describe the complexity of identifying target mutations and changes, they do highlight current support for some form of human genome editing, specifically for correcting harmful or deleterious mutations. Further analysis of these issues is inevitable as research on using multiomics to uncover molecular determinants of human cognitive performance advances and as insights from these studies are increasingly applied to performance modification—whether for restoring abilities or enhancing them—because this research probes the aspects or brain biology that makes individuals their unique selves.

___________________

Potential Future Directions

Multiomics combined with advanced data analytics and computational techniques offers a window into the molecular determinants underlying cognition and behavior. However, given the complexity of brain function, establishing a link between these traits and molecular profiles involves a better understanding of brain circuits and how they interact within the brain under various conditions and over time. More recent research has used multiomics to study the link between gut microbiome and cognitive function and disorders (Li et al., 2022; Liang et al., 2022; Foster and Trivedi, 2024). How microbiome composition influences the brain and cognitive abilities in the context of health and disease is a relatively new area of research (NASEM, 2023a). Determining whether modification of the microbiome could be harnessed for modification of cognitive abilities requires more exploration. As noted earlier, neural imaging (e.g., with functional magnetic resonance imaging, or fMRI) can provide unique insights into the brain that allow for linking of patterns of brain activity to cognitive behavior. However, standardizing neuroimaging data across large populations of individuals to establish a reliable link can be challenging. Furthermore, imaging systems can differ across various research and healthcare components, and their respective data often cannot be easily shared and compared on large scales in the same way as genomic and multiomic information (Giordano, 2012). These challenges multiply for multimodel analyses of brain function that seek to integrate imaging data with text-based data from multiomics studies and physio-electrical data from electroencephalogram (EEG) measurements (e.g., electrome) (De Loof, 2016). Finally, advances in computational tools, ML, and AI will likely enable multimodal analyses in the future.