Deep learning techniques can recognize complex patterns in noisy, multidimensional data. In recent years, researchers have started to explore the potential of deep learning in the field of structural biology, including protein crystallography. This field has some significant challenges, in particular producing high-quality and well ordered protein crystals. Additionally, collecting diffraction data with high completeness and quality, and determining and refining protein structures can be problematic. Protein crystallographic data are often high-dimensional, noisy and incomplete. Deep learning algorithms can extract relevant features from these data and learn to recognize patterns, which can improve the success rate of crystallization and the quality of crystal structures. This paper reviews progress in this field.

Ubiquitin ligases regulate core cellular processes through diverse mechanisms. The ubiquitin ligase COP1 is conserved from plants to humans and is particularly important for targeting developmental transcription factors for ubiquitination. COP1 can function independently, but can also be recruited to Cullin-4 ubiquitin ligase complexes via the DET1 adaptor protein. However, the mechanism of action of complexes containing COP1 and DET1 is not well understood. Here we report the cryo-electron microscopy structure of human DET1, bound to proteins that enable Cullin-4 recruitment (DDB1-DDA1) and an additional ubiquitin ligase enzyme (Ube2e2). We observe that DDA1 stabilises a closed conformation of DET1, binding adjacent to a unique Ube2e2 binding-insert in DET1. Moreover, we demonstrate that closure of DET1 underlies COP1 recruitment, which binds in an antiparallel dimeric state. Disrupting either the Ube2e2-binding insertion of DET1, or distinct recruitment sites on COP1, abolish DET1–COP1 binding and DET1-mediated modulation of COP1 levels. The multifaceted architecture provides an efficient platform for ubiquitination of substrates, or COP1 itself, by Cullin-4DET1 and offers multiple opportunities for physiological regulation. ### Competing Interest Statement The authors have declared no competing interest.

IntroductionProteins that adopt multiple conformations pose significant challenges in structural biology research and pharmaceutical development, as structure determination via single particle cryo-electron microscopy (cryo-EM) is often impeded by data heterogeneity. In this context, the enhanced signal-to-noise ratio of single molecule cryo-electron diffraction (simED) offers a promising alternative. However, a significant challenge in diffraction methods is the loss of phase information, which is crucial for accurate structure determination.MethodsHere, we present DiffraGAN, a conditional generative adversarial network (cGAN) that estimates the missing phases at high resolution from a combination of single particle high-resolution diffraction data and low-resolution image data.ResultsFor simulated datasets, DiffraGAN allows effectively determining protein structures at atomic resolution from diffraction patterns and noisy low-resolution images.DiscussionOur findings suggest that combining single particle cryo-electron diffraction with advanced generative modeling, as in DiffraGAN, could revolutionize the way protein structures are determined, offering an alternative and complementary approach to existing methods.

As an alternative approach to X-ray crystallography and single-particle cryo-electron microscopy, single-molecule electron diffraction has a better signal-to-noise ratio and the potential to increase the resolution of protein models. This technology requires collection of numerous diffraction patterns, which can lead to congestion of data collection pipelines. However, only a minority of the diffraction data are useful for structure determination because the chances of hitting a protein of interest with a narrow electron beam may be small. This necessitates novel concepts for quick and accurate data selection. For this purpose, a set of machine learning algorithms for diffraction data classification has been implemented and tested. The proposed pre-processing and analysis workflow efficiently distinguished between amorphous ice and carbon support, providing proof of the principle of machine learning based identification of positions of interest. While limited in its current context, this approach exploits inherent characteristics of narrow electron beam diffraction patterns and can be extended for protein data classification and feature extraction.

Multiple studies have established associations between human gut bacteria and host physiology, but determining the molecular mechanisms underlying these associations has been challenging 1 – 3 . Akkermansia muciniphila has been robustly associated with positive systemic effects on host metabolism, favourable outcomes to checkpoint blockade in cancer immunotherapy and homeostatic immunity 4 – 7 . Here we report the identification of a lipid from A. muciniphila ’s cell membrane that recapitulates the immunomodulatory activity of A. muciniphila in cell-based assays 8 . The isolated immunogen, a diacyl phosphatidylethanolamine with two branched chains (a15:0-i15:0 PE), was characterized through both spectroscopic analysis and chemical synthesis. The immunogenic activity of a15:0-i15:0 PE has a highly restricted structure–activity relationship, and its immune signalling requires an unexpected toll-like receptor TLR2–TLR1 heterodimer 9 , 10 . Certain features of the phospholipid’s activity are worth noting: it is significantly less potent than known natural and synthetic TLR2 agonists; it preferentially induces some inflammatory cytokines but not others; and, at low doses (1% of EC 50 ) it resets activation thresholds and responses for immune signalling. Identifying both the molecule and an equipotent synthetic analogue, its non-canonical TLR2–TLR1 signalling pathway, its immunomodulatory selectivity and its low-dose immunoregulatory effects provide a molecular mechanism for a model of A. muciniphila’ s ability to set immunological tone and its varied roles in health and disease.

Significance The protein kinase MEKK1 activates stress-signaling pathways in response to various cellular stressors, including chemotherapies that disrupt dynamics of the tubulin cytoskeleton. We show that MEKK1 contains a previously uncharacterized domain that can preferentially bind to the curved tubulin heterodimer—which is found in soluble tubulin and at sites of microtubule assembly and disassembly. Mutations that interfere with MEKK1−tubulin binding disrupt microtubule networks in migrating cells and are enriched in patient-derived tumor sequences. These results suggest that MEKK1−tubulin binding may be relevant to cancer progression, and the efficacy of microtubule-disrupting chemotherapies that require the activity of MEKK1.

The MEKK1 protein is a pivotal kinase activator of responses to cellular stress. Activation of MEKK1 can trigger various responses, including mitogen activated protein (MAP) kinases, NF-κB signalling, or cell migration. Notably, MEKK1 activity is triggered by microtubule-targeting chemotherapies, amongst other stressors. Here we show that MEKK1 contains a previously unidentified tumour overexpressed gene (TOG) domain. The MEKK1 TOG domain binds to tubulin heterodimers—a canonical function of TOG domains—but is unusual in that it appears alone rather than as part of a multi-TOG array, and has structural features distinct from previously characterised TOG domains. MEKK1 TOG demonstrates a clear preference for binding curved tubulin heterodimers, which exist in soluble tubulin and at sites of microtubule polymerisation and depolymerisation. Mutations disrupting tubulin-binding lead to destabilisation of the MEKK1 protein in cells, and ultimately a decrease in microtubule density at the leading edge of polarised cells. We also show that MEKK1 mutations at the tubulin-binding interface of the TOG domain recur in patient derived tumour sequences, suggesting selective enrichment of tumour cells with disrupted MEKK1–microtubule association. Together, these findings provide a direct link between the MEKK1 protein and tubulin, which is likely to be relevant to cancer cell migration and response to microtubule-modulating therapies.SIGNIFICANCE STATEMENT The protein kinase MEKK1 activates stress response pathways in response to various cellular stressors, including chemotherapies that disrupt dynamics of the tubulin cytoskeleton. Filipčík et al., show that MEKK1 contains a previously uncharacterised domain that can preferentially bind to the curved tubulin heterodimer—which is found at sites of microtubule assembly and disassembly. Mutations that interfere with MEKK1-tubulin binding disrupt microtubule networks in migrating cells and are enriched in patient-derived tumour sequences. These results suggest that MEKK1-tubulin binding may be relevant to cancer progression, and the efficacy of microtubule-disrupting chemotherapies that require the activity of MEKK1.

Phosphorylation and ubiquitination are pervasive post-translational modifications that impact all processes inside eukaryotic cells. The role of each modification has been studied for decades, and functional interplay between the two has long been demonstrated and even more widely postulated. However, our understanding of the molecular features that allow phosphorylation to control protein ubiquitination and ubiquitin to control phosphorylation has only recently begun to build. Here, we review examples of regulation between ubiquitination and phosphorylation, aiming to describe mechanisms at the molecular level. In general, these examples illustrate phosphorylation as a versatile switch throughout ubiquitination pathways, and ubiquitination primarily impacting kinase signalling in a more emphatic manner through scaffolding or degradation. Examples of regulation between these two processes are likely to grow even further as advances in molecular biology, proteomics, and computation allow a system-level understanding of signalling. Many new cases could involve similar principles to those described here, but the extensive co-regulation of these two systems leaves no doubt that they still have many surprises in store.

S100P belongs to the S100 family of calcium-binding proteins regulating diverse cellular processes. Certain S100 family members (S100A4 and S100B) are associated with cancer and used as biomarkers of metastatic phenotype. Also S100P is abnormally expressed in tumors and implicated in migration-invasion, survival, and response to therapy. Here we show that S100P binds the tumor suppressor protein p53 as well as its negative regulator HDM2, and that this interaction perturbs the p53-HDM2 binding and increases the p53 level. Paradoxically, the S100P-induced p53 is unable to activate its transcriptional targets hdm2, p21WAF, and bax following the DNA damage. This appears to be related to reduced phosphorylation of serine residues in both N-terminal and C-terminal regions of the p53 molecule. Furthermore, the S100P expression results in lower levels of pro-apoptotic proteins, in reduced cell death response to cytotoxic treatments, followed by stimulation of therapy-induced senescence and increased clonogenic survival. Conversely, the S100P silencing suppresses the ability of cancer cells to survive the DNA damage and form colonies. Thus, we propose that the oncogenic role of S100P involves binding and inactivation of p53, which leads to aberrant DNA damage responses linked with senescence and escape to proliferation. Thereby, the S100P protein may contribute to the outgrowth of aggressive tumor cells resistant to cytotoxic therapy and promote cancer progression.