Central nervous system tumours represent one of the most lethal cancer types, particularly among children¹. Primary treatment includes neurosurgical resection of the tumour, in which a delicate balance must be struck between maximizing the extent of resection and minimizing risk of neurological damage and comorbidity^2,3. However, surgeons have limited knowledge of the precise tumour type prior to surgery. Current standard practice relies on preoperative imaging and intraoperative histological analysis, but these are not always conclusive and occasionally wrong. Using rapid nanopore sequencing, a sparse methylation profile can be obtained during surgery⁴. Here we developed Sturgeon, a patient-agnostic transfer-learned neural network, to enable molecular subclassification of central nervous system tumours based on such sparse profiles. Sturgeon delivered an accurate diagnosis within 40 minutes after starting sequencing in 45 out of 50 retrospectively sequenced samples (abstaining from diagnosis of the other 5 samples). Furthermore, we demonstrated its applicability in real time during 25 surgeries, achieving a diagnostic turnaround time of less than 90 min. Of these, 18 (72%) diagnoses were correct and 7 did not reach the required confidence threshold. We conclude that machine-learned diagnosis based on low-cost intraoperative sequencing can assist neurosurgical decision-making, potentially preventing neurological comorbidity and avoiding additional surgeries.

Background: Nanopore-based DNA sequencing relies on basecalling the electric current signal. Basecalling requires neural networks to achieve competitive accuracies. To improve sequencing accuracy further, new models are continuously proposed with new architectures. However, benchmarking is currently not standardized, and evaluation metrics and datasets used are defined on a per publication basis, impeding progress in the field. This makes it impossible to distinguish data from model driven improvements. Results: To standardize the process of benchmarking, we unified existing benchmarking datasets and defined a rigorous set of evaluation metrics. We benchmarked the latest seven basecaller models by recreating and analyzing their neural network architectures. Our results show that overall Bonito’s architecture is the best for basecalling. We find, however, that species bias in training can have a large impact on performance. Our comprehensive evaluation of 90 novel architectures demonstrates that different models excel at reducing different types of errors and using recurrent neural networks (long short-term memory) and a conditional random field decoder are the main drivers of high performing models. Conclusions: We believe that our work can facilitate the benchmarking of new basecaller tools and that the community can further expand on this work.

Redox signaling is controlled by the reversible oxidation of cysteine thiols, a post-translational modification triggered by H₂O₂ acting as a second messenger. However, H₂O₂ actually reacts poorly with most cysteine thiols and it is not clear how H₂O₂ discriminates between cysteines to trigger appropriate signaling cascades in the presence of dedicated H₂O₂ scavengers like peroxiredoxins (PRDXs). It was recently suggested that peroxiredoxins act as peroxidases and facilitate H₂O₂-dependent oxidation of redox-regulated proteins via disulfide exchange reactions. It is unknown how the peroxiredoxin-based relay model achieves the selective substrate targeting required for adequate cellular signaling. Using a systematic mass-spectrometry-based approach to identify cysteine-dependent interactors of the five human 2-Cys peroxiredoxins, we show that all five human 2-Cys peroxiredoxins can form disulfide-dependent heterodimers with a large set of proteins. Each isoform displays a preference for a subset of disulfide-dependent binding partners, and we explore isoform-specific properties that might underlie this preference. We provide evidence that peroxiredoxin-based redox relays can proceed via two distinct molecular mechanisms. Altogether, our results support the theory that peroxiredoxins could play a role in providing not only reactivity but also selectivity in the transduction of peroxide signals to generate complex cellular signaling responses.

Besides genetic alterations, the cellular environment also determines disease onset and progression. When different cell types contribute to disease outcome, this imposes environmental challenges as different cell types likely differ in their extracellular dependencies. Hsa‐microRNA‐31‐5p (miR‐31) is highly expressed in keratinocytes of psoriatic skin, and we show that expression in keratinocytes is induced by limited glucose availability and enables increased survival under limiting glucose conditions by increasing glutamine metabolism. In addition, miR‐31 expression results in not only secretion of specific metabolites (aspartate and glutamate) but also secretion of immunomodulatory factors. We show that this miR‐31‐induced secretory phenotype is sufficient to induce Th17 cell differentiation, a hallmark of psoriasis. Inhibitors of miR31‐induced metabolic rewiring and metabolic crosstalk with immune cells alleviate psoriasis pathology in a mouse model of psoriasis. Together our data illustrate an emerging concept of metabolic interaction across cell compartments that characterizes disease development, which can be employed to design effective treatment options for disease, as shown here for psoriasis.

Genomic DNA is under constant oxidative damage, with 8-oxo-7,8-dihydro-2’-deoxyguanosine (8-oxo-dG) being the prominent lesion linked to mutagenesis, epigenetics, and gene regulation. Existing methods to detect 8-oxo-dG rely on indirect approaches, while nanopore sequencing enables direct detection of base modifications. A model for 8-oxo-dG detection is currently missing due to the lack of training data. Here, we develop a strategy using synthetic oligos to generate long, 8-oxo-dG context-variable DNA molecules for deep learning and nanopore sequencing. Our training approach addresses the rarity of 8-oxo-dG relative to guanine, enabling specific detection. Applied to a tissue culture model of oxidative damage, our method reveals uneven genomic 8-oxo-dG distribution, dissimilar context pattern to C>A mutations, and local 5-mC depletion. This dual measurement of 5-mC and 8-oxo-dG at single-molecule resolution uncovers new insights into their interplay. Our approach also provides a general framework for detecting other rare DNA modifications using synthetic DNA and nanopore sequencing.