The FeatureCloud Platform for Federated Learning in Biomedicine: Unified Approach.

BACKGROUND: Machine learning and artificial intelligence have shown promising results in many areas and are driven by the increasing amount of available data. However, these data are often distributed across different institutions and cannot be easily shared owing to strict privacy regulations. Federated learning (FL) allows the training of distributed machine learning models without sharing sensitive data. In addition, the implementation is time-consuming and requires advanced programming skills and complex technical infrastructures. OBJECTIVE: Various tools and frameworks have been developed to simplify the development of FL algorithms and provide the necessary technical infrastructure. Although there are many high-quality frameworks, most focus only on a single application case or method. To our knowledge, there are no generic frameworks, meaning that the existing solutions are restricted to a particular type of algorithm or application field. Furthermore, most of these frameworks provide an application programming interface that needs programming knowledge. There is no collection of ready-to-use FL algorithms that are extendable and allow users (eg, researchers) without programming knowledge to apply FL. A central FL platform for both FL algorithm developers and users does not exist. This study aimed to address this gap and make FL available to everyone by developing FeatureCloud, an all-in-one platform for FL in biomedicine and beyond. METHODS: The FeatureCloud platform consists of 3 main components: a global frontend, a global backend, and a local controller. Our platform uses a Docker to separate the local acting components of the platform from the sensitive data systems. We evaluated our platform using 4 different algorithms on 5 data sets for both accuracy and runtime. RESULTS: FeatureCloud removes the complexity of distributed systems for developers and end users by providing a comprehensive platform for executing multi-institutional FL analyses and implementing FL algorithms. Through its integrated artificial intelligence store, federated algorithms can easily be published and reused by the community. To secure sensitive raw data, FeatureCloud supports privacy-enhancing technologies to secure the shared local models and assures high standards in data privacy to comply with the strict General Data Protection Regulation. Our evaluation shows that applications developed in FeatureCloud can produce highly similar results compared with centralized approaches and scale well for an increasing number of participating sites. CONCLUSIONS: FeatureCloud provides a ready-to-use platform that integrates the development and execution of FL algorithms while reducing the complexity to a minimum and removing the hurdles of federated infrastructure. Thus, we believe that it has the potential to greatly increase the accessibility of privacy-preserving and distributed data analyses in biomedicine and beyond.

Longitudinal, Multi-Platform Metagenomics Yields a High-Quality Genomic Catalog and Guides an In Vitro Model for Cheese Communities.

Microbiomes are intricately intertwined with human health, geochemical cycles, and food production. While many microbiomes of interest are highly complex and experimentally intractable, cheese rind microbiomes have proven to be powerful model systems for the study of microbial interactions. To provide a more comprehensive view of the genomic potential and temporal dynamics of cheese rind communities, we combined longitudinal, multi-platform metagenomics of three ripening washed-rind cheeses with whole-genome sequencing of community isolates. Sequencing-based approaches revealed a highly reproducible microbial succession in each cheese and the coexistence of closely related Psychrobacter species and enabled the prediction of plasmid and phage diversity and their host associations. In combination with culture-based approaches, we established a genomic catalog and a paired 16-member in vitro washed-rind cheese system. The combination of multi-platform metagenomic time-series data and an in vitro model provides a rich resource for further investigation of cheese rind microbiomes both computationally and experimentally. IMPORTANCE Metagenome sequencing can provide great insights into microbiome composition and function and help researchers develop testable hypotheses. Model microbiomes, such as those composed of cheese rind bacteria and fungi, allow the testing of these hypotheses in a controlled manner. Here, we first generated an extensive longitudinal metagenomic data set. This data set reveals successional dynamics, yields a phyla-spanning bacterial genomic catalog, associates mobile genetic elements with their hosts, and provides insights into functional enrichment of Psychrobacter in the cheese environment. Next, we show that members of the washed-rind cheese microbiome lend themselves to in vitro community reconstruction. This paired metagenomic data and in vitro system can thus be used as a platform for generating and testing hypotheses related to the dynamics within, and the functions associated with, cheese rind microbiomes.

Native mass spectrometry-based metabolomics identifies metal-binding compounds.

Although metals are essential for the molecular machineries of life, systematic methods for discovering metal-small molecule complexes from biological samples are limited. Here, we describe a two-step native electrospray ionization-mass spectrometry method, in which post-column pH adjustment and metal infusion are combined with ion identity molecular networking, a rule-based data analysis workflow. This method enabled the identification of metal-binding compounds in complex samples based on defined mass (m/z) offsets of ion species with the same chromatographic profiles. As this native electrospray metabolomics approach is suited to the use of any liquid chromatography-mass spectrometry system to explore the binding of any metal, this method has the potential to become an essential strategy for elucidating metal-binding molecules in biology.

Experimental approaches to tracking mobile genetic elements in microbial communities.

Horizontal gene transfer is an important mechanism of microbial evolution and is often driven by the movement of mobile genetic elements between cells. Due to the fact that microbes live within communities, various mechanisms of horizontal gene transfer and types of mobile elements can co-occur. However, the ways in which horizontal gene transfer impacts and is impacted by communities containing diverse mobile elements has been challenging to address. Thus, the field would benefit from incorporating community-level information and novel approaches alongside existing methods. Emerging technologies for tracking mobile elements and assigning them to host organisms provide promise for understanding the web of potential DNA transfers in diverse microbial communities more comprehensively. Compared to existing experimental approaches, chromosome conformation capture and methylome analyses have the potential to simultaneously study various types of mobile elements and their associated hosts. We also briefly discuss how fermented food microbiomes, given their experimental tractability and moderate species complexity, make ideal models to which to apply the techniques discussed herein and how they can be used to address outstanding questions in the field of horizontal gene transfer in microbial communities.

A Proposed Chaperone of the Bacterial Type VI Secretion System Functions To Constrain a Self-Identity Protein.

The bacterium Proteus mirabilis can communicate identity through the secretion of the self-identity protein IdsD via the type VI secretion (T6S) system. IdsD secretion is essential for self-versus-nonself recognition behaviors in these populations. Here we provide an answer to the unresolved question of how the activity of a T6S substrate, such as IdsD, is regulated before secretion. We demonstrate that IdsD is found in clusters that form independently of the T6S machinery and activity. We show that the IdsC protein, which is a member of the proposed DUF4123 chaperone family, is essential for the maintenance of these clusters and of the IdsD protein itself. We provide evidence that amino acid disruptions in IdsC are sufficient to disrupt IdsD secretion but not IdsD localization into subcellular clusters, strongly supporting the notion that IdsC functions in at least two different ways: maintaining IdsD levels and secreting IdsD. We propose that IdsC, and likely other DUF4123-containing proteins, functions to regulate T6S substrates in the donor cell both by maintaining protein levels and by mediating secretion at the T6S machinery.IMPORTANCE Understanding the subcellular dynamics of self-identity proteins is crucial for developing models of self-versus-nonself recognition. We directly addressed how a bacterium restricts self-identity information before cell-cell exchange. We resolved two conflicting models for type VI secretion (T6S) substrate regulation by focusing on the self-identity protein IdsD. One model is that a cognate immunity protein binds the substrate, inhibiting activity before transport. Another model proposes that DUF4123 proteins act as chaperones in the donor cell, but no detailed molecular mechanism was previously known. We resolve this discrepancy and propose a model wherein a chaperone couples IdsD sequestration with its localization. Such a molecular mechanism restricts the communication of identity, and possibly other T6S substrates, in producing cells.

A single point mutation in a TssB/VipA homolog disrupts sheath formation in the type VI secretion system of Proteus mirabilis.

The type VI secretion (T6S) system is a molecular device for the delivery of proteins from one cell into another. T6S function depends on the contractile sheath comprised of TssB/VipA and TssC/VipB proteins. We previously reported on a mutant variant of TssB that disrupts T6S-dependent export of the self-identity protein, IdsD, in the bacterium Proteus mirabilis. Here we determined the mechanism underlying that initial observation. We show that T6S-dependent export of multiple self-recognition proteins is abrogated in this mutant strain. We have mapped the mutation, which is a single amino acid change, to a region predicted to be involved in the formation of the TssB-TssC sheath. We have demonstrated that this mutation does indeed inhibit sheath formation, thereby explaining the global disruption of T6S activity. We propose that this mutation could be utilized as an important tool for studying functions and behaviors associated with T6S systems.

The Self-Identity Protein IdsD Is Communicated between Cells in Swarming Proteus mirabilis Colonies.

Proteus mirabilis is a social bacterium that is capable of self (kin) versus nonself recognition. Swarming colonies of this bacterium expand outward on surfaces to centimeter-scale distances due to the collective motility of individual cells. Colonies of genetically distinct populations remain separate, while those of identical populations merge. Ids proteins are essential for this recognition behavior. Two of these proteins, IdsD and IdsE, encode identity information for each strain. These two proteins bind in vitro in an allele-restrictive manner. IdsD-IdsE binding is correlated with the merging of populations, whereas a lack of binding is correlated with the separation of populations. Key questions remained about the in vivo interactions of IdsD and IdsE, specifically, whether IdsD and IdsE bind within single cells or whether IdsD-IdsE interactions occur across neighboring cells and, if so, which of the two proteins is exchanged. Here we demonstrate that IdsD must originate from another cell to communicate identity and that this nonresident IdsD interacts with IdsE resident in the recipient cell. Furthermore, we show that unbound IdsD in recipient cells does not cause cell death and instead appears to contribute to a restriction in the expansion radius of the swarming colony. We conclude that P. mirabilis communicates IdsD between neighboring cells for nonlethal kin recognition, which suggests that the Ids proteins constitute a type of cell-cell communication. IMPORTANCE: We demonstrate that self (kin) versus nonself recognition in P. mirabilis entails the cell-cell communication of an identity-encoding protein that is exported from one cell and received by another. We further show that this intercellular exchange affects swarm colony expansion in a nonlethal manner, which adds social communication to the list of potential swarm-related regulatory factors.