Deep learning of yeast protein production efficiencies

Biosustainability concept, where carbon resources are recycled rather than exploited, represents an important economic driving force in the coming decades. Microbial cell factories that are able to utilize industrial by-products (like CO2) or low-value sugars and efficiently produce a variety of chemicals will play a major role in this process, where yeast is one of the preferred host organisms. To make those processes cost effective, we need to understand the cellular regulation mechanisms, where protein production is one of the key elements. Previously collected data on mRNA and protein abundances indicate >1000-fold differences in protein production rates [1]. Understanding these complex regulatory mechanisms using deep learning algorithms would allow one to create novel more efficient platform strains for sustainable production of chemicals [2].

Our aim is to create a deep learning algorithm for the yeast genomic data to understand the regulation of protein production based on the experimental data published in Lahtvee et al1.

Work description:
•\tAnalyzing large-scale transcriptomic and proteomic datasets from literature1.
•\tDeveloping a deep learning platform for phenotype predictions2.
•\tPredicting cause-effect relationships for the protein production in yeast.
•\tModel validation based on the experimental data.

1 Lahtvee PJ, Sánchez BJ, […], Nielsen J (2017) Absolute quantification of protein and mRNA abundances demonstrate variability in gene-specific translation efficiency in yeast. Cell Systems 4:495-504.e5.
2 Märtens, K., Hallin, J., Warringer, J., Liti, G., & Parts, L. (2016). Predicting quantitative traits from genome and phenome with near perfect accuracy. Nature Communications, 7, 11512.
Graduation Theses defence year
Leopold Parts, Petri-Jan Lahtvee
Spoken language (s)
Estonian, English
Requirements for candidates

Application of contact

Leopold Parts