»Understanding the genetic regulatory code governing gene expression and being able to control it is a primary challenge in molecular biology. To this end, data-driven machine learning can be used to predict gene expression patterns, such as mRNA abundance, directly from nucleotide sequence. Deep neural networks automatically learn informative sequence representations and interpreting them enables us to improve our understanding of gene regulation.

I will present the application of deep learning to uncover the genetic regulatory code controlling mRNA abundance in model organisms ranging from bacteria to plants. In all organisms, mRNA abundance can be predicted directly from DNA sequence, which encodes up to 82% of the variation of transcript levels. Analysis of regulatory motifs demonstrates that it is not single motifs or regions, but the entire gene regulatory structure spanning coding and non-coding regulatory regions as well as the correct combination of regulatory elements that define gene expression levels.

By coupling the above genomic regulatory principles with generative adversarial networks (GAN), a novel strategy to design de novo synthetic regulatory DNA can be devised. The method captures the whole regulatory sequence-expression landscape in a gene-specific manner, generating regulatory DNA with prespecified target mRNA levels. As the designs span the entire gene regulatory structure and are highly divergent from natural DNA, the method outperforms existing approaches and demonstrates the usefulness of generative modeling for sequence design.«

Vljudno vabljeni! Seminar bo potekal v angleškem jeziku. / You are cordially invited to attend this lecture, which will be delivered in  English.

Dodatne informacije: dr. Jan Zrimec, jan.zrimec@nib.si.