biology · health · agriculture · environment · family: the answer is in a field you’ve never heard of
thegenomeisn'ttheblueprint
We still cannot predict an organism's traits from its genome
Problem statement
Decades after the first genome was sequenced, the biggest gap in biological knowledge remains our inability to look at an organism's genetics and environment and predict its observable characteristics (phenotype). NSF has designated "Understanding the Rules of Life: Predicting Phenotype" as one of its formal Big Ideas for Future Investments. The theoretical constructs that explain and predict characteristics of living systems — from molecular components to cells, whole organisms, communities, and biomes — remain largely undiscovered. This is not a data quantity problem: we have massive genomic datasets. It is a knowledge gap about how genetic information interacts with epigenetic regulation, environmental context, and emergent properties across levels of biological organization to produce phenotypes.
Why this matters
Genotype-to-phenotype prediction would transform medicine (predicting disease risk and drug response from genome), agriculture (designing crops for specific environments), conservation (predicting which species can adapt to climate change), and biotechnology (engineering organisms with desired properties). Current approaches to crop improvement, personalized medicine, and conservation genetics rely largely on statistical associations rather than mechanistic understanding — associations that can break down in new genetic backgrounds, new environments, or new species. For plants specifically, NSF's Plant Genome Research Program says breakthroughs in tools, technologies, and resources are still needed to understand how plant genomes, and their interactions with other genomes and with the environment, give rise to the traits of the world's plants.
What’s been tried and why it hasn’t worked
Functional genomics has focused narrowly on candidate genes and single-gene manipulation for genes of large effect, which captures only a fraction of phenotypic variation — many traits are controlled by many genes of small effect. Genome-wide association studies (GWAS) identify statistical associations but not causal mechanisms, and predictions often do not transfer across populations or environments. Epigenetic complexity — heritable phenotypic properties that occur without genome sequence modification — has fundamentally complicated the picture beyond what sequence-based approaches can capture. Emergent properties arise from complex, nonlinear interactions among biological systems that in isolation do not exhibit such properties — determining these emergent network properties is a critical unsolved problem. For plants, transformation bottlenecks (genotype-dependent regeneration from tissue culture) are a major bottleneck for gene-function research and crop improvement in many plant species.
What would unlock progress
Systems-level analyses of gene-regulatory networks and their emergent functional properties, elucidating causal connections across levels of biological organization. High-throughput phenotyping tools linked to genomic data, especially for non-model organisms and field conditions. AI/ML approaches capable of integrating multi-omic datasets (genomic, proteomic, metabolomic, phenomic) into predictive frameworks. Genotype-independent plant transformation methods that circumvent tissue culture bottlenecks. Integrative approaches bridging plant physiology, ecology, evolution, and development through engineering and quantitative modeling.
Entry points for student teams
A student team could select a specific quantitative trait in a model organism (e.g., drought tolerance in Arabidopsis) and compare the predictive accuracy of different modeling approaches — GWAS, gene network models, machine learning on multi-omic data — using publicly available datasets from TAIR or Phytozome, identifying where each approach fails and why. Alternatively, a team could develop a high-throughput phenotyping protocol using computer vision and low-cost imaging hardware for a specific crop trait (plant height, leaf area, color-based stress indicators) in a campus greenhouse. Relevant skills: genomics, bioinformatics, machine learning, plant biology, image processing.
Genes
Structural cousins: stuck the same way, in other fields
Sources
"Understanding the Rules of Life: Predicting Phenotype," NSF Big Ideas; "IOS Core Programs," NSF 24-546 (IntBIO Track); "PGRP: Plant Genome Research Program," NSF 24-547; "URoL:ASC," NSF 23-512. (accessed 2026-02-15). go to source ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
- "Understanding the Rules of Life" is one of NSF's 10 Big Ideas for Future Investments, reflecting the fundamental importance and difficulty of this problem.
- Cross-domain connection: shares structure with `bio-organism-climate-response-prediction` (predicting organismal response to climate requires genotype-phenotype prediction in environmental context) and `bio-ecological-forecasting-skill-gap` (ecological forecasting is limited by our inability to predict how individual organisms will respond).
- The `failure:not-attempted` tag applies because the multi-scale, multi-omic integration required for genuine genotype-to-phenotype prediction has not been attempted at the systems level for most organisms — work has focused on individual genes or statistical associations rather than mechanistic, predictive models.
- The `failure:disciplinary-silo` tag applies because genomics, phenomics, ecology, and systems biology remain largely separate fields with different data types, tools, and conceptual frameworks.
- The IntBIO track within IOS funds collaborative proposals that span sub-disciplines within biology and use experiment, modeling, and advanced computation to understand how biological systems work across scales.
- Note reconciled 2026-08-20: a note above argues for `failure:not-attempted`; the genome now carries `failure:theoretical-gap`, `failure:disciplinary-silo` after a taxonomy revision. The original note is kept verbatim as the tagging rationale of record.
Reconciliation 2026-09-29: A site review read this write-up against the four NSF documents on its Source line. All were fetched on 2026-09-29. NSF 24-547 (Plant Genome Research Program) was read at the cited address and at its solicitation page, https://www.nsf.gov/funding/opportunities/pgrp-plant-genome-research-program/5338/nsf24-547/solicitation. NSF 24-546 (Division of Integrative Organismal Systems Core Programs) was read at https://www.nsf.gov/publications/pub_summ.jsp?ods_key=nsf24546, and NSF 23-512 (Using the Rules of Life to Address Societal Challenges) at https://www.nsf.gov/publications/pub_summ.jsp?ods_key=nsf23512. The Big Idea page was read in its current form at https://www.nsf.gov/about/history/big-ideas and in Internet Archive copies of https://www.nsf.gov/news/special_reports/big_ideas/life.jsp dated 2018 to 2023. Several passages in this write-up were not in those four documents. Some of them were traced to four other NSF documents, fetched the same day: NSF 20-512 (Understanding the Rules of Life: Epigenetics) at https://www.nsf.gov/publications/pub_summ.jsp?ods_key=nsf20512, NSF 21-560 (Understanding the Rules of Life: Emergent Networks) at https://www.nsf.gov/publications/pub_summ.jsp?ods_key=nsf21560, NSF 20-532 (Enabling Discovery through GEnomic Tools, EDGE) at https://www.nsf.gov/publications/pub_summ.jsp?ods_key=nsf20532, and NSF 23-019 (Advancing Plant Transformation, a Dear Colleague Letter that NSF 24-547 names) at https://www.nsf.gov/publications/pub_summ.jsp?ods_key=nsf23019. These are not on the Source line. Changes: (1) The text said "we have petabytes of genomic data." No document gives a size. NSF 24-547 speaks of "massive" datasets, and the sentence now says that. (2) The text said "most traits are influenced by hundreds or thousands of loci with small effects." No document gives those counts. NSF 20-532 says many traits are "controlled by many genes of small effect," and the sentence now says that. (3) The text said transformation bottlenecks "prevent functional validation in most crop species." NSF 23-019 says that for many plant species regeneration depends heavily on genotype, and that this is "a major bottleneck for fundamental research and crop improvement." The sentence now follows that wording. (4) The text said that plant species "with complex genomes, difficult breeding systems, and long generation times cannot be improved using the statistical approaches that work for model organisms, leaving most of the world's crop diversity inaccessible to genomics-assisted improvement." None of the documents says this. The sentence was replaced with what NSF 24-547 says about plants. (5) The text said current approaches "all rely on statistical associations" that "break down" in new settings. None of the documents says this. It was softened to "rely largely" and "can break down." (6) The text said predictions from genome-wide association studies "do not transfer across populations or environments." None of the documents mentions these studies. It was softened to "often do not transfer." (7) A note above said the IntBIO track "explicitly funds proposals that integrate biology with engineering, math, and physical sciences to address the rules of life." NSF 24-546 says IntBIO proposals must span sub-disciplines within the Biological Sciences directorate, and its description of the track does not mention engineering, mathematics, physical sciences or the rules of life. The note now follows the solicitation. Confirmed and left as written: the four titles and numbers on the Source line (the 2018 archived page carries the title "Understanding the Rules of Life: Predicting Phenotype"); the statement that the inability to predict phenotype from genetics and environment is the biggest gap in biological knowledge (current Big Idea page); the 10 Big Ideas for future NSF investment; the sentence on theoretical constructs and the description of epigenetics (NSF 20-512); the sentence on emergent properties (NSF 21-560); the focus on candidate genes and single genes of large effect, and the call for systems-level analysis of gene-regulatory networks (NSF 20-532); high-throughput phenotyping, machine learning, transformation methods that do not depend on genotype or tissue culture, and the bridging of plant physiology, ecology, evolution and development (NSF 24-547). Not found in any document fetched, and left as written because they are general statements without figures: the list of uses in medicine, agriculture, conservation and biotechnology; that genome-wide association studies find statistical links and not causes; and the words "field conditions" in the phenotyping sentence. Items 5 and 6 remain without a source after softening. NOT checked: whether the TAIR and Phytozome datasets named in the entry points are open without a subscription or login, because both sites load only in a browser. NSF now marks 24-547 and 24-546 as archived. Genes untouched.