AI intern for plants genetics

At Neoplants, R&D team is constantly interested in developing new techniques & technologies.

As AI intern for plants genetics at Neoplants, you will work on the development of an AI model to predict the best genetic scaffold for a given metabolic pathway in Nicotiana benthamiana.

Context:

The optimum expression of a complex metabolic pathway in a plant is currently a challenge as it is influenced by numerous factors such as the genes sequences, mRNA stability, codon bias, protein folding, stability and post translational modification. But the first elements that come into play are the genetic regulatory elements, such as promoters and terminators. For the model plant Nicotiana benthamiana there are an extensive number of publications describing the relative efficiencies of such regulatory elements. However these elements are always tested in a specific context which can affect the results and always in low complexity context (1 or 2 genes).

At Neoplants, we design complex metabolic pathways with more than ten genes to improve plants. We then move these pathways into plants that scientists know little about, adding more complexity. To make the pathway work well, we have to try many different regulatory elements and figure out the right combination through trial and error. If we don’t get it right, even one poorly expressed gene can reduce how well the entire pathway works.

This is where an AI model could drastically increase the speed of iteration of our products.

Objective:

To overcome this limited amount of information our strategy is to develop an AI model that would be able to predict the set of genetic regulatory elements to use to reach the desired expression of complex metabolic pathways.

• Identify the best available foundation models to use

• Find and/or develop the right embedding for our purpose

• Training and fine tuning of the model to predict a gene expression level based on a set of genetic regulatory elements.

• Increase the complexity of the genetic scaffold to more and more complex scaffold (from one gene to N genes)

• Ultimately train on various plant species in order to be able to predict the best scaffold not only in N. benthamiana but in any given plant.

• Fetch experimental data and genetic elements metadata from our LIMS (Laboratory Information and Management System), Benchling.

You will also get the chance to discover the working environment of an R&D lab in a dynamic and vibrant startup.

Project Title: AI model to predict the best genetic scaffold

Project Leader: Jérôme Bodin