Accurate structure prediction of biomolecular interactions with AlphaFold 3.

The introduction of AlphaFold 21 has spurred a revolution in modelling the structure of proteins and their interactions, enabling a huge range of applications in protein modelling and design2-6. Here we describe our AlphaFold 3 model with a substantially updated diffusion-based architecture that is capable of predicting the joint structure of complexes including proteins, nucleic acids, small molecules, ions and modified residues. The new AlphaFold model demonstrates substantially improved accuracy over many previous specialized tools: far greater accuracy for protein-ligand interactions compared with state-of-the-art docking tools, much higher accuracy for protein-nucleic acid interactions compared with nucleic-acid-specific predictors and substantially higher antibody-antigen prediction accuracy compared with AlphaFold-Multimer v.2.37,8. Together, these results show that high-accuracy modelling across biomolecular space is possible within a single unified deep-learning framework.

Polarizable coarse-grained models for molecular dynamics simulation of liquid cyclohexane.

Force field parameters for polarizable coarse-grained (CG) supra-atomic models of liquid cyclohexane are proposed. Two different bead sizes were investigated, one representing two fine-grained (FG) CH(2)r united atoms of the cyclohexane ring, and one representing three FG CH(2)r united atoms. Electronic polarizability is represented by a massless charge-on-spring particle connected to each CG bead. The model parameters were calibrated against the experimental density and heat of vaporization of liquid cyclohexane, and the free energy of cyclohexane hydration. Both models show good agreement with thermodynamic properties of cyclohexane, yet overestimate the self-diffusion. The dielectric properties of the polarizable models agree very well with experiment.

Identification of ALP+/CD73+ defining markers for enhanced osteogenic potential in human adipose-derived mesenchymal stromal cells by mass cytometry.

BACKGROUND: The impressive progress in the field of stem cell research in the past decades has provided the ground for the development of cell-based therapy. Mesenchymal stromal cells obtained from adipose tissue (AD-MSCs) represent a viable source for the development of cell-based therapies. However, the heterogeneity and variable differentiation ability of AD-MSCs depend on the cellular composition and represent a strong limitation for their use in therapeutic applications. In order to fully understand the cellular composition of MSC preparations, it would be essential to analyze AD-MSCs at single-cell level. METHOD: Recent advances in single-cell technologies have opened the way for high-dimensional, high-throughput, and high-resolution measurements of biological systems. We made use of the cytometry by time-of-flight (CyTOF) technology to explore the cellular composition of 17 human AD-MSCs, interrogating 31 markers at single-cell level. Subcellular composition of the AD-MSCs was investigated in their naïve state as well as during osteogenic commitment, via unsupervised dimensionality reduction as well as supervised representation learning approaches. RESULT: This study showed a high heterogeneity and variability in the subcellular composition of AD-MSCs upon isolation and prolonged culture. Algorithm-guided identification of emerging subpopulations during osteogenic differentiation of AD-MSCs allowed the identification of an ALP+/CD73+ subpopulation of cells with enhanced osteogenic differentiation potential. We could demonstrate in vitro that the sorted ALP+/CD73+ subpopulation exhibited enhanced osteogenic potential and is moreover fundamental for osteogenic lineage commitment. We finally showed that this subpopulation was present in freshly isolated human adipose-derived stromal vascular fractions (SVFs) and that could ultimately be used for cell therapies. CONCLUSION: The data obtained reveal, at single-cell level, the heterogeneity of AD-MSCs from several donors and highlight how cellular composition impacts the osteogenic differentiation capacity. The marker combination (ALP/CD73) can not only be used to assess the differentiation potential of undifferentiated AD-MSC preparations, but also could be employed to prospectively enrich AD-MSCs from the stromal vascular fraction of human adipose tissue for therapeutic applications.

Author Correction: Automated Gleason grading of prostate cancer tissue microarrays via deep learning.

A correction to this article has been published and is linked from the HTML and PDF versions of this paper. The error has not been fixed in the paper.

GM-CSF and CXCR4 define a T helper cell signature in multiple sclerosis.

Cytokine dysregulation is a central driver of chronic inflammatory diseases such as multiple sclerosis (MS). Here, we sought to determine the characteristic cellular and cytokine polarization profile in patients with relapsing-remitting multiple sclerosis (RRMS) by high-dimensional single-cell mass cytometry (CyTOF). Using a combination of neural network-based representation learning algorithms, we identified an expanded T helper cell subset in patients with MS, characterized by the expression of granulocyte-macrophage colony-stimulating factor and the C-X-C chemokine receptor type 4. This cellular signature, which includes expression of very late antigen 4 in peripheral blood, was also enriched in the central nervous system of patients with relapsing-remitting multiple sclerosis. In independent validation cohorts, we confirmed that this cell population is increased in patients with MS compared with other inflammatory and non-inflammatory conditions. Lastly, we also found the population to be reduced under effective disease-modifying therapy, suggesting that the identified T cell profile represents a specific therapeutic target in MS.

Automated Gleason grading of prostate cancer tissue microarrays via deep learning.

The Gleason grading system remains the most powerful prognostic predictor for patients with prostate cancer since the 1960s. Its application requires highly-trained pathologists, is tedious and yet suffers from limited inter-pathologist reproducibility, especially for the intermediate Gleason score 7. Automated annotation procedures constitute a viable solution to remedy these limitations. In this study, we present a deep learning approach for automated Gleason grading of prostate cancer tissue microarrays with Hematoxylin and Eosin (H&E) staining. Our system was trained using detailed Gleason annotations on a discovery cohort of 641 patients and was then evaluated on an independent test cohort of 245 patients annotated by two pathologists. On the test cohort, the inter-annotator agreements between the model and each pathologist, quantified via Cohen's quadratic kappa statistic, were 0.75 and 0.71 respectively, comparable with the inter-pathologist agreement (kappa = 0.71). Furthermore, the model's Gleason score assignments achieved pathology expert-level stratification of patients into prognostically distinct groups, on the basis of disease-specific survival data available for the test cohort. Overall, our study shows promising results regarding the applicability of deep learning-based solutions towards more objective and reproducible prostate cancer grading, especially for cases with heterogeneous Gleason patterns.

Similarities and Differences of Blood N-Glycoproteins in Five Solid Carcinomas at Localized Clinical Stage Analyzed by SWATH-MS.

Cancer is mostly incurable when diagnosed at a metastatic stage, making its early detection via blood proteins of immense clinical interest. Proteomic changes in tumor tissue may lead to changes detectable in the protein composition of circulating blood plasma. Using a proteomic workflow combining N-glycosite enrichment and SWATH mass spectrometry, we generate a data resource of 284 blood samples derived from patients with different types of localized-stage carcinomas and from matched controls. We observe whether the changes in the patient's plasma are specific to a particular carcinoma or represent a generic signature of proteins modified uniformly in a common, systemic response to many cancers. A quantitative comparison of the resulting N-glycosite profiles discovers that proteins related to blood platelets are common to several cancers (e.g., THBS1), whereas others are highly cancer-type specific. Available proteomics data, including a SWATH library to study N-glycoproteins, will facilitate follow-up biomarker research into early cancer detection.

Sensitive detection of rare disease-associated cell subsets via representation learning.

Rare cell populations play a pivotal role in the initiation and progression of diseases such as cancer. However, the identification of such subpopulations remains a difficult task. This work describes CellCnn, a representation learning approach to detect rare cell subsets associated with disease using high-dimensional single-cell measurements. Using CellCnn, we identify paracrine signalling-, AIDS onset- and rare CMV infection-associated cell subsets in peripheral blood, and extremely rare leukaemic blast populations in minimal residual disease-like situations with frequencies as low as 0.01%.