RESEARCH NOTE 002 / RESEARCH DIRECTION
Where the next signal could come from
The scientific foothold is real: specific DNA changes can alter development. The open challenge is to find rules that transfer beyond one well-studied example.
Our early predictors have not established an advantage beyond strong biological controls. That result points toward more precise questions: a particular structure, an independently recurring evolutionary change, or a measured difference in developmental activity.
Evidence that the direction is grounded
In published mouse experiments, Kvon and colleagues showed that replacing a limb enhancer with snake versions could severely reduce limb development. Altered versions could restore enhancer function. This connects particular regulatory sequences to an anatomical outcome in an experimental setting. It does not mean a single enhancer explains all snake anatomy. Kvon et al., Cell, 2016.
Other researchers have found repeated loss or inactivation of gastric genes in independently evolved stomachless fish lineages. Such comparisons link anatomical change to genomic change, while leaving open whether some gene losses caused an organ to disappear or followed its loss. Kato et al., Communications Biology, 2024.
These are discoveries by the cited investigators. They support the research direction. DNALYZER has not independently reproduced their biological experiments or discovered a new causal anatomical rule.
Three paths worth testing
This order reflects our judgment about an exploratory program using existing public data. It is not a set of calibrated breakthrough probabilities.
Learn from repeated structural changes
Study a well-defined anatomical function across lineages that changed independently. Test whether genomic differences add information beyond shared ancestry, sequence quality, and annotation gaps.
Success would mean: a specific association transfers to an evolutionary transition kept out of discovery. A missing annotation must not be mistaken for a missing gene; correlation must not be mistaken for a cause.
Predict changes in developmental activity
Use public experiments that compare regulatory sequences in a defined tissue and developmental context. Predict the change in measured activity between sequences, with related constructs and entire loci kept together during evaluation.
Success would mean: improvement on untouched loci or studies, supported by independent checks. Recognizing an active enhancer is a different task from predicting what a sequence change will do. An assay in one context does not establish whole-organism construction.
Give each genome a richer organism model
Represent structures, counts, dimensions, developmental stages, and uncertainty as linked, sourced assertions. Preserve which structures are comparable across species and when an annotation is unknown or inapplicable.
Success would mean: richer, independently checked phenotype information improves a predefined anatomical task beyond coarse labels. An LLM can help extract and reconcile evidence; its expectations about a species are not observations.
What would actually be new?
Pairing rich anatomy with genomes is established prior art. “Reverse Genomics” already connected mammalian phenomic characters with evolutionary loss of conserved noncoding elements. Our contribution would need to be a demonstrably better resource, a more reliable predictive test, or a previously unreported biological relationship. Marcovitz et al., Molecular Biology and Evolution, 2016.
Public genomic data and LLMs make it easier to assemble and test hypotheses. They cannot supply missing measurements, eliminate shared ancestry, or turn a model’s confident explanation into a causal finding.
A founding program with decision points
- Establish a trustworthy benchmark. Confirm the anatomical definitions, source context, usable sequence coverage, and independence of the comparisons.
- Recover a known result. Verify that the workflow can reproduce a published association before searching for new ones.
- Run a held-out discovery test. Compare against strong baselines, preserve failed results, and seek independent reproduction.
- Match the claim to the evidence. A new association is progress. A mechanistic claim needs correspondingly stronger support.
This is enough to justify a focused exploratory team spanning comparative genomics, developmental biology, and scientific engineering. It is not yet evidence for a universal DNA decoder. The first milestone should be a narrow result that survives scrutiny.
Sources and useful starting points
- Kvon et al. (2016) — functional investigation of limb-enhancer evolution.
- Kato et al. (2024) — convergent gene losses in stomachless fishes.
- Marcovitz et al. (2016), Reverse Genomics — direct prior art for rich phenotype–genome mapping.
- VISTA Enhancer Browser — public developmental enhancer experiments.
- Hiller laboratory tools and data — public comparative genomics resources.
Publication policy: these notes share aggregate findings, source attribution, and limitations. Internal implementation and future candidate work are not included. Public summaries alone are insufficient to reproduce every analysis.
Explore the founding mission