A central chaperone-like role for 14-3-3 proteins in human cells.

14-3-3 proteins are highly conserved regulatory proteins that interact with hundreds of structurally diverse clients and act as central hubs of signaling networks. However, how 14-3-3 paralogs differ in specificity and how they regulate client protein function are not known for most clients. Here, we map the interactomes of all human 14-3-3 paralogs and systematically characterize the effect of disrupting these interactions on client localization. The loss of 14-3-3 binding leads to the coalescence of a large fraction of clients into discrete foci in a client-specific manner, suggesting a central chaperone-like function for 14-3-3 proteins. Congruently, the engraftment of 14-3-3 binding motifs to nonclients can suppress their aggregation or phase separation. Finally, we show that 14-3-3s negatively regulate the localization of the RNA-binding protein SAMD4A to cytoplasmic granules and inhibit its activity as a translational repressor. Our work suggests that 14-3-3s have a more prominent role as chaperone-like molecules than previously thought.

Comprehensive interactome profiling of the human Hsp70 network highlights functional differentiation of J domains.

Hsp70s comprise a deeply conserved chaperone family that has a central role in maintaining protein homeostasis. In humans, Hsp70 client specificity is provided by 49 different co-factors known as J domain proteins (JDPs). However, the cellular function and client specificity of JDPs have largely remained elusive. We have combined affinity purification-mass spectrometry (AP-MS) and proximity-dependent biotinylation (BioID) to characterize the interactome of all human JDPs and Hsp70s. The resulting network suggests specific functions for many uncharacterized JDPs, and we establish a role of conserved JDPs DNAJC9 and DNAJC27 in histone chaperoning and ciliogenesis, respectively. Unexpectedly, we find that the J domain of DNAJC27 but not of other JDPs can fully replace the function of endogenous DNAJC27, suggesting a previously unappreciated role for J domains themselves in JDP specificity. More broadly, our work expands the role of the Hsp70-regulated proteostasis network and provides a platform for further discovery of JDP-dependent functions.

Batch equalization with a generative adversarial network.

MOTIVATION: Advances in automation and imaging have made it possible to capture a large image dataset that spans multiple experimental batches of data. However, accurate biological comparison across the batches is challenged by batch-to-batch variation (i.e. batch effect) due to uncontrollable experimental noise (e.g. varying stain intensity or cell density). Previous approaches to minimize the batch effect have commonly focused on normalizing the low-dimensional image measurements such as an embedding generated by a neural network. However, normalization of the embedding could suffer from over-correction and alter true biological features (e.g. cell size) due to our limited ability to interpret the effect of the normalization on the embedding space. Although techniques like flat-field correction can be applied to normalize the image values directly, they are limited transformations that handle only simple artifacts due to batch effect. RESULTS: We present a neural network-based batch equalization method that can transfer images from one batch to another while preserving the biological phenotype. The equalization method is trained as a generative adversarial network (GAN), using the StarGAN architecture that has shown considerable ability in style transfer. After incorporating new objectives that disentangle batch effect from biological features, we show that the equalized images have less batch information and preserve the biological information. We also demonstrate that the same model training parameters can generalize to two dramatically different types of cells, indicating this approach could be broadly applicable. AVAILABILITY AND IMPLEMENTATION: https://github.com/tensorflow/gan/tree/master/tensorflow_gan/examples/stargan. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Pervasive mislocalization of pathogenic coding variants underlying human disorders.

Widespread sequencing has yielded thousands of missense variants predicted or confirmed as disease-causing. This creates a new bottleneck: determining the functional impact of each variant - largely a painstaking, customized process undertaken one or a few genes or variants at a time. Here, we established a high-throughput imaging platform to assay the impact of coding variation on protein localization, evaluating 3,547 missense variants of over 1,000 genes and phenotypes. We discovered that mislocalization is a common consequence of coding variation, affecting about one-sixth of all pathogenic missense variants, all cellular compartments, and recessive and dominant disorders alike. Mislocalization is primarily driven by effects on protein stability and membrane insertion rather than disruptions of trafficking signals or specific interactions. Furthermore, mislocalization patterns help explain pleiotropy and disease severity and provide insights on variants of unknown significance. Our publicly available resource will likely accelerate the understanding of coding variation in human diseases.

ECNet is an evolutionary context-integrated deep learning framework for protein engineering.

Machine learning has been increasingly used for protein engineering. However, because the general sequence contexts they capture are not specific to the protein being engineered, the accuracy of existing machine learning algorithms is rather limited. Here, we report ECNet (evolutionary context-integrated neural network), a deep-learning algorithm that exploits evolutionary contexts to predict functional fitness for protein engineering. This algorithm integrates local evolutionary context from homologous sequences that explicitly model residue-residue epistasis for the protein of interest with the global evolutionary context that encodes rich semantic and structural features from the enormous protein sequence universe. As such, it enables accurate mapping from sequence to function and provides generalization from low-order mutants to higher-order mutants. We show that ECNet predicts the sequence-function relationship more accurately as compared to existing machine learning algorithms by using ~50 deep mutational scanning and random mutagenesis datasets. Moreover, we used ECNet to guide the engineering of TEM-1 ß-lactamase and identified variants with improved ampicillin resistance with high success rates.