De novo protein design for catalysis of any desired chemical reaction is a long standing goal in protein engineering, due to the broad spectrum of technological, scientific and medical applications. Currently, mapping protein sequence to protein function is, however, neither computationionally nor experimentally tangible 1,2 . Here we developed ProteinGAN, a specialised variant of the generative adversarial network 3 that is able to 'learn' natural protein sequence diversity and enables the generation of functional protein sequences. ProteinGAN learns the evolutionary relationships of protein sequences directly from the complex multidimensional amino acid sequence space and creates new, highly diverse sequence variants with natural-like physical properties. Using malate dehydrogenase as a template enzyme, we show that 24% of the ProteinGAN-generated and experimentally tested sequences are soluble and display wild-type level catalytic activity in the tested conditions in vitro , even in highly mutated (>100 mutations) sequences.ProteinGAN therefore demonstrates the potential of artificial intelligence to rapidly generate highly diverse novel functional proteins within the allowed biological constraints of the sequence space.

Comparison of conventional and newly developed laccase-based electrocatalytic systems for oxygen reduction.

De novo protein design for catalysis of any desired chemical reaction is a long standing goal in protein engineering, due to the broad spectrum of technological, scientific and medical applications. Currently, mapping protein sequence to protein function is, however, neither computationionally nor experimentally tangible 1,2 . Here we developed ProteinGAN, a specialised variant of the generative adversarial network 3 that is able to 'learn' natural protein sequence diversity and enables the generation of functional protein sequences. ProteinGAN learns the evolutionary relationships of protein sequences directly from the complex multidimensional amino acid sequence space and creates new, highly diverse sequence variants with natural-like physical properties. Using malate dehydrogenase as a template enzyme, we show that 24% of the ProteinGAN-generated and experimentally tested sequences are soluble and display wild-type level catalytic activity in the tested conditions in vitro , even in highly mutated (>100 mutations) sequences.ProteinGAN therefore demonstrates the potential of artificial intelligence to rapidly generate highly diverse novel functional proteins within the allowed biological constraints of the sequence space.

No abstract

Human activating signal cointegrator homology (ASCH) domain-containing proteins are widespread and diverse but, at present, the vast majority of those proteins have no function assigned to them. This study demonstrates that the 103-amino acid Escherichia coli protein YqfB, previously identified as hypothetical, is a unique ASCH domain-containing amidohydrolase responsible for the catabolism of N 4-acetylcytidine (ac4C). YqfB has several interesting and unique features: i) it is the smallest monomeric amidohydrolase described to date, ii) it is active towards structurally different N 4-acylated cytosines/cytidines, and iii) it has a high specificity for these substrates (k cat /K m up to 2.8 × 10 6 M −1 s −1). Moreover, our results suggest that YqfB contains a unique Thr-Lys-Glu catalytic triad, and Arg acting as an oxyanion hole. The mutant lacking the yqfB gene retains the ability to grow, albeit poorly, on N 4-acetylcytosine as a source of uracil, suggesting that an alternative route for the utilization of this compound exists in E. coli. Overall, YqfB ability to hydrolyse various N 4-acylated cytosines and cytidines not only sheds light on the long-standing mystery of how ac4C is catabolized in bacteria, but also expands our knowledge of the structural diversity within the active sites of amidohydrolases. Human activating signal cointegrator 1 (ASC-1), otherwise known as the thyroid hormone receptor interactor protein 4 (ASC-1/TRIP4), interacts with a wide range of unrelated transcription factors to facilitate nuclear receptors-mediated transcription. It also plays a pivotal role in the transactivation of serum response factor (SRF), activating protein 1 (AP-1), and nuclear factor κB (NF-κB) 1,2. In 2006 it was shown that the C-terminal domain of ASC-1 defines a large ASC-1 homology (ASCH) domain superfamily 2. To date, the ASCH-containing proteins have been reported for a wide range of organisms representing all three kingdoms of life, and have also been found in viruses 3. Although it has long been suggested that this domain of ~110 residues may be responsible for RNA binding during transcription coactivation, RNA processing, and regulation of translation, the vast majority of ASCH proteins are small (~140 residues) hypothetical proteins, which at present have no function assigned to them 2. One such protein is the product of the yqfB gene in Escherichia coli. Herein we demonstrate that the 103-amino acid YqfB is a unique monomeric amidohydrolase responsible for the catabolism of the modified nucleoside, N 4-acetylcytidine (ac4C), in E. coli. More than 160 of differently modified nucleotides play a crucial role in various biological processes 1,2,4,5. The biosynthetic pathways of many modified bases, nucleosides and nucleotides are well understood 5,6 , but the catabolism or salvage of those compounds are scarcely studied. Just like the ASCH proteins, the modified nucleoside ac4C is found in organisms within all three domains of life 4-9. It prevents misreading of AUA isoleucine codons during protein synthes...