New Genetic Creations

This project explores the use of machine learning to generate grayscale images based on the hox genes of various organisms. Hox genes are essential in controlling the body plan of animals and other eukaryotes, and their sequences exhibit patterns unique to each species. By using hox genes and corresponding images, this project aims to explore the hidden patterns that life's creation is based on.

---

The development of genome and gene sequencing technologies was a milestone in biology, providing unprecedented insight into the genetic material of living organisms. With this breakthrough came the promise of unlocking the secrets of life, from understanding how genes interact with each other to finding cures for many diseases. However, as we delved deeper into the genetic complexity of life, it became clear that the interplay between genes is much more intricate and convoluted than initially anticipated. Despite significant progress in identifying genes and their functions, there is still much we do not know about how these genetic building blocks work together to form complex biological systems.


Today, interdisciplinary teams of scientists, including bioinformaticians, biochemists, geneticists, and medical researchers, work together to unravel the complex system of how genes interact with one another. Reductionism has been a prevalent view in the scientific community, where the aim is to understand complex phenomena by breaking them down into smaller, more manageable components. However, this approach has limitations, as it fails to capture the emergent properties that arise from the interactions between these components. In this project, I aim to explore the intersection between reductionism and emergence in the context of genetic interplay, and the use of machine learning models as a tool to bridge the gap between these two perspectives. When studying the black box of genetic interaction, it is only fitting to approach it with another black box - a machine learning model. In particular, the machine learning model ‘Stable Diffusion’ has demonstrated promising results when it comes to creating impressive visual representations of text-based data.


For this project, Hox gene sequences were obtained from the National Center for Biotechnology Information (NCBI). The sequences were truncated to a 158-character string, as this is the maximum input for the ‘Stable Diffusion‘ model. Images of the corresponding species were selected and edited into greyscale images, which function as a 2D representation for the body of these organisms. An embedding on the pre-trained stable diffusion model version 1.5 was trained on 80 images with its corresponding DNA.


Through this project, I hope to expand my understanding of gene interactions and ultimately create representations of new organisms. Questioning how does life work? Are we living in a simulation? What are the patterns behind the creation of life?

While the initial prototype utilizes a 158 character string of the hox genes, the training on the complete sets of hox genes of these species is planned in the future. Furthermore, experimental testing with new generated hox gene strings is also planned. Ultimately, instead of 2D grayscale images, I want to generate 3D models based on hox genes.

At the moment, I'm working on an installation that presents the images that result from the audience's live DNA input, playing with the notion of creative creation and nature’s process of creation.