Fast and reliable detection of patients with severe and heterogeneous illnesses is a major goal of precision medicine^1,2. Patients with leukaemia can be identified using machine learning on the basis of their blood transcriptomes³. However, there is an increasing divide between what is technically possible and what is allowed, because of privacy legislation^4,5. Here, to facilitate the integration of any medical data from any data owner worldwide without violating privacy laws, we introduce Swarm Learning—a decentralized machine-learning approach that unites edge computing, blockchain-based peer-to-peer networking and coordination while maintaining confidentiality without the need for a central coordinator, thereby going beyond federated learning. To illustrate the feasibility of using Swarm Learning to develop disease classifiers using distributed data, we chose four use cases of heterogeneous diseases (COVID-19, tuberculosis, leukaemia and lung pathologies). With more than 16,400 blood transcriptomes derived from 127 clinical studies with non-uniform distributions of cases and controls and substantial study biases, as well as more than 95,000 chest X-ray images, we show that Swarm Learning classifiers outperform those developed at individual sites. In addition, Swarm Learning completely fulfils local confidentiality regulations by design. We believe that this approach will notably accelerate the introduction of precision medicine.

Background: Alpha-synuclein is a key protein implicated in the pathogenesis of Parkinson's disease (PD). It is the main component of the Lewy bodies, a cardinal neuropathological feature in the disease. In addition, whole locus multiplications and point mutations in the gene coding for alpha-synuclein lead to autosomal dominant monogenic PD. Over the past decade, research on PD has impelled the development of new animal models based on alpha-synuclein. In this context, transgenic mouse lines have failed to reproduce several hallmarks of PD, especially the strong and progressive dopaminergic neurodegeneration over time that occurs in the patients. In contrast, viral vector-based models in rats and non-human primates display prominent, although highly variable, nigral dopaminergic neuron loss. However, the few studies available on viral vector-mediated overexpression of alpha-synuclein in mice report a weak neurodegenerative process and no clear Lewy body-like pathology. To address this issue, we performed a comprehensive comparative study of alpha-synuclein overexpression by means of recombinant adeno-associated viral vectors serotype 2/7 (rAAV2/7) at different doses in adult mouse substantia nigra.Results: We noted a significant and dose-dependent alpha-synucleinopathy over time upon nigral viral vector-mediated alpha-synuclein overexpression. We obtained a strong, progressive and dose-dependent loss of dopaminergic neurons in the substantia nigra, reaching a maximum of 82% after 8 weeks. This effect correlated with a reduction in tyrosine hydroxylase immunoreactivity in the striatum. Moreover, behavioural analysis revealed significant motor impairments from 12 weeks after injection on. In addition, we detected the presence of alpha-synuclein-positive aggregates in the remaining surviving neurons. When comparing wild-type to mutant A53T alpha-synuclein at the same vector dose, both induced a similar degree of cell death. These data were supported by a biochemical analysis that showed a net increase in soluble and insoluble alpha-synuclein expression over time to the same extent for both alpha-synuclein variants.Conclusions: In conclusion, ourin vivodata provide evidence that strong and significant alpha-synuclein-induced neuropathology and progressive dopaminergic neurodegeneration can be achieved in mouse brain by means of rAAV2/7.

Conversion of soluble α-synuclein into insoluble and fibrillar inclusions is a hallmark of Parkinson’s disease and other synucleinopathies. Accumulating evidence points towards a relationship between its generation at nerve terminals and structural synaptic pathology. Little is known about the pathogenic impact of α-synuclein conversion and deposition at nigrostriatal dopaminergic synapses in transgenic mice, mainly owing to expression limitations of the α-synuclein construct. Here, we explore whether both the rat as a model and expression of the bacterial artificial chromosome construct consisting of human full-length wild-type α-synuclein could exert dopaminergic neuropathological effects. We found that the human promoter induced a pan-neuronal expression, matching the rodent α-synuclein expression pattern, however, with prominent C-terminally truncated fragments. Ageing promoted conversion of both full-length and C-terminally truncated α-synuclein species into insolube and proteinase K-resistant fibres, with strongest accumulation in the striatum, resembling biochemical changes seen in human Parkinson’s disease. Transgenic rats develop early changes in novelty-seeking, avoidance and smell before the progressive motor deficit. Importantly, the observed pathological changes were associated with severe loss of the dopaminergic integrity, thus resembling more closely the human pathology.

Lewy bodies, a pathological hallmark of Parkinson's disease (PD), contain aggregated alpha-synuclein (αSyn), which is found in several modified forms and can be discovered phosphorylated, ubiquitinated and truncated. Aggregation-prone truncated species of αSyn caused by aberrant cleavage of this fibrillogenic protein are hypothesized to participate in its sequestration into inclusions subsequently leading to synaptic dysfunction and neuronal death. Here, we investigated the role of calpain cleavage of αSyn in vivo by generating two opposing mouse models. We crossed into human [A30P]αSyn transgenic (i) mice deficient for calpastatin, a calpain-specific inhibitor, thus enhancing calpain activity (SynCAST(−)) and (ii) mice overexpressing human calpastatin leading to reduced calpain activity (SynCAST(+)). As anticipated, a reduced calpain activity led to a decreased number of αSyn-positive aggregates, whereas loss of calpastatin led to increased truncation of αSyn in SynCAST(−). Furthermore, overexpression of calpastatin decreased astrogliosis and the calpain-dependent degradation of synaptic proteins, potentially ameliorating the observed neuropathology in [A30P]αSyn and SynCAST(+) mice. Overall, our data further support a crucial role of calpains, particularly of calpain 1, in the pathogenesis of PD and in disease-associated aggregation of αSyn, indicating a therapeutic potential of calpain inhibition in PD.

Testing of new therapeutic strategies for Parkinson's disease (PD) is currently hampered by the lack of relevant and reproducible animal models. Here, we developed a robust rat model for PD by injection of adeno-associated viral vectors (rAAV2/7) encoding α-synuclein into the substantia nigra, resulting in reproducible nigrostriatal pathology and behavioral deficits in a 4-week time period. Progressive dopaminergic dysfunction was corroborated by histopathologic and biochemical analysis, motor behavior testing and in vivo microdialysis. L-DOPA treatment was found to reverse the behavioral phenotype. Non-invasive positron emission tomography imaging and magnetic resonance spectroscopy allowed longitudinal monitoring of neurodegeneration. In addition, insoluble α-synuclein aggregates were formed in this model. This α-synuclein rat model shows improved face and predictive validity, and therefore offers the possibility to reliably test novel therapeutics. Furthermore, it will be of great value for further research into the molecular pathogenesis of PD and the importance of α-synuclein aggregation in the disease process.

Purpose Genome sequencing (GS) is expected to reduce the diagnostic gap in rare disease genetics. We aimed to evaluate a scalable framework for genome-based analyses ‘beyond the exome’ in regular care of patients with inherited retinal degeneration (IRD) or inherited optic neuropathy (ION). Methods PCR-free short-read GS was performed on 1000 consecutive probands with IRD/ION in routine diagnostics. Complementary whole-blood RNA-sequencing (RNA-seq) was done in a subset of 74 patients. An open-source bioinformatics analysis pipeline was optimised for structural variant (SV) calling and combined RNA/DNA variation interpretation. Results A definite genetic diagnosis was established in 57.4% of cases. For another 16.7%, variants of uncertain significance were identified in known IRD/ION genes, while the underlying genetic cause remained unresolved in 25.9%. SVs or alterations in non-coding genomic regions made up for 12.7% of the observed variants. The RNA-seq studies supported the classification of two unclear variants. Conclusion GS is feasible in clinical practice and reliably identifies causal variants in a substantial proportion of individuals. GS extends the diagnostic yield to rare non-coding variants and enables precise determination of SVs. The added diagnostic value of RNA-seq is limited by low expression levels of the major IRD disease genes in blood.