Inspiration
Current synthetic biology tools lack the capability to realistically reconstruct full prehistoric genomes at chromosome-level resolution. Genomancer bridges this gap by combining AI-driven comparative genomics with exhaustive genome synthesis to enable scientifically plausible organism design based on phenotypic descriptions. On a personal note, this story is grounded in a deep fascination with prehistoric life since childhood.
What it does
Genomancer transforms written phenotypic traits into:
- A fully assembled, biologically justified karyotype.
- Complete AGCT sequences for each chromosome.
- System-specific gene annotations and interactive genome maps.
- Realistic, genome-derived regulatory network models.
- A beyond-theoretical, step-by-step pathway for synthetic assembly and incubation of extinct species, conditional on ethical clearance.
How we built it
I integrated:
- AI models trained on extant avian and reptilian genomic data.
- A chromosome segmentation system for realistic genome assembly.
- Full-sequence generators producing exhaustive AGCT output per chromosome.
- Gene annotation logic based on modular biological systems (e.g., neural, skeletal, metabolic).
- Network modeling derived directly from genome structure.
- A guided theoretical organism construction pipeline.
- UI components with genome-wide interactive visualization and export capabilities.
Challenges we ran into
Memory Overload: Trying to generate and render massive sequences (200+ million base pairs) Blocking Operations: Synchronous sequence generation was freezing the UI Component Conflicts: Multiple sequence viewer components with conflicting logic.
Accomplishments that we're proud of
- Achieved full-chromosome, sequence-accurate genome generation entirely from phenotypic text.
- Developed system-specific gene annotation for transparent genome exploration.
- Constructed realistic, non-redundant regulatory networks.
- Delivered a technically grounded organism assembly pathway that moves beyond speculative theory.
What we learned
Through building Genomancer, I demonstrated that exhaustive, full-chromosome genome reconstruction is achievable within AI-first, single-prompt app architectures. I learned that realistic regulatory network modeling cannot rely on generic templates -- it requires direct inference from the generated genome itself to preserve biological plausibility. Additionally, I found that clear genome visualization, chromosome-specific logic, and transparent construction pathways significantly enhance usability, making the platform accessible to both synthetic biology experts and enthusiasts.
What's next for Genomancer
The next steps are grounded in a bold but realistic vision: laying the scientific and technical foundation for what could one day become the world’s first responsible, mission-driven “Jurassic Park” - not as entertainment, but as a pioneering biotech venture at the intersection of synthetic biology, evolutionary science, and conservation. To move in that direction, I will deepen the platform’s biological fidelity by integrating probabilistic genome variation to realistically simulate evolutionary divergence and genetic diversity. I also plan to expand our regulatory network modeling with advanced epigenetic inference to accurately capture gene expression control mechanisms. Collaborating closely with leading synthetic biology experts, I aim to assess the true experimental feasibility of the theoretical organism construction pathways generated by Genomancer. At every stage, I am committed to exploring the legal, ethical, and ecological frameworks necessary to ensure that any real-world application of this technology is guided by responsibility, transparency, and scientific integrity.
Log in or sign up for Devpost to join the conversation.