Tobias Sikosek

Background: Alzheimer's disease (AD), the most common form of dementia, is typically diagnosed using the AT(N) framework, which assesses amyloid, tau, and neurodegeneration biomarkers. However, this framework has limitations, particularly in predicting cognitive decline among amyloid‐positive individuals. Method: This study explores the synergistic potential of combining small RNA biomarkers with established protein markers to improve the early detection of AD. Using data from 1,913 participants aged 50+ from the European Prevention of Alzheimer's Dementia (EPAD) clinical trial, all free of dementia at enrollment, we performed ultra‐deep small RNA sequencing on whole blood samples. A refined sequencing protocol reduced erythroid RNA interference, enabling the identification of rare biomarker RNAs. We define high and low amyloid groups based on a cutoff on the p ‐tau ₁₈₁ /Ab _1‐42 ratio as determined from cerebrospinal fluid. Result: Analysis revealed small RNAs that predicted early cognitive decline (Clinical Dementia Rating of 0.5) with an AUC of ∼0.7, which increased to 0.77 within the high‐amyloid subgroup. Functional evaluation linked these RNAs to dementia‐relevant pathways, including neuronal, cardiovascular, and inflammatory activities. Conclusion: Our findings highlight that combining small RNA and amyloid protein markers significantly outperforms either biomarker type alone, offering a more robust and precise approach to early AD detection. This integrated strategy addresses current gaps in the AT(N) framework, enhancing stratification and enabling more effective interventions. Small nucleolar RNAs and microRNAs emerge as promising candidates for further exploration as part of this synergistic diagnostic model.

The characterisation of circulating nucleic acid biomarkers in liquid biopsy based diagnostics holds great potential to transform the landscape of early cancer detection and screening. These tests are increasingly incorporating information beyond the primary sequence, to include epigenetic, fragmentomic, and other chemical properties to boost performance. Chemical modifications of RNA offer a rich, and currently underutilised source of biomarker signal. We aimed to develop a single-molecule method to profile 2'-O-methylation of a ribosomal RNA fragment previously linked to lung cancer, and to evaluate its diagnostic value in blood. We have designed a targeted capture strategy that ligates structure-guided adapters to a ~22-nucleotide ribosomal RNA fragment to enable sequencing of native molecules on a nanopore platform. Raw ionic-current signals were used to train and validate machine-learning classifiers to detect methylation states of synthetic oligonucleotides and cell culture derived ribosomal RNA fragments. Finally, we applied these methods to liquid biopsy samples collected from a 43-patient cohort of individuals undergoing investigation for suspected lung cancer. Here we show that single molecules of the target fragment are sequenced, and their methylation states can be accurately (92%) and quantitatively (Pearson r = 0.997) measured. In clinical liquid biopsy samples, it reveals a differential pattern of methylation in lung cancer that yields a diagnostic classifier with an area under the receiver-operating characteristic curve of 0.84. This approach enables direct, single-molecule methylation profiling of small RNAs in blood and identifies a lung cancer-associated methylation pattern with diagnostic potential. It is readily compatible with multi-modal liquid biopsy assays to enhance performance.

Small RNAs hold crucial biological information and have immense diagnostic and therapeutic value. While many established annotation tools focus on microRNAs, there are myriads of other small RNAs that are currently underutilized. These small RNAs can be difficult to annotate, as ground truth is limited and well-established mapping and mismatch rules are lacking. TransfoRNA is a machine learning framework based on Transformers that explores an alternative strategy. It uses common annotation tools to generate a small seed of high-confidence training labels, while then expanding upon those labels iteratively. TransfoRNA learns sequence-specific representations of all RNAs to construct a similarity network which can be interrogated as new RNAs are annotated, allowing to rank RNAs based on their familiarity. While models can be flexibly trained on any RNA dataset, we here present a version trained on TCGA (The Cancer Genome Atlas) small RNA sequences and demonstrate its ability to add annotation confidence to an unrelated dataset, where 21% of previously unannotated RNAs could be annotated. Relative to its training data, TransfoRNA could boost high-confidence annotations in TCGA by ∼50% while providing transparent explanations even for low-confidence ones. It could learn to annotate 97% of isomiRs from just single examples and confidently identify new members of other familiar classes with high accuracy, while reliably rejecting false RNAs. All source code is available at https://github.com/gitHBDX/TransfoRNA and can be executed at Code Ocean ( https://codeocean.com/capsule/5415298/ ). An interactive website is available at www.transforna.com . GRAPHICAL

Immunotherapies have recently gained traction as highly effective therapies in a subset of late-stage cancers. Unfortunately, only a minority of patients experience the remarkable benefits of immunotherapies, whilst others fail to respond or even come to harm through immune-related adverse events. For immunotherapies within the PD-1/PD-L1 inhibitor class, patient stratification is currently performed using tumor (tissue-based) PD-L1 expression. However, PD-L1 is an accurate predictor of response in only ~30% of cases. There is pressing need for more accurate biomarkers for immunotherapy response prediction. We sought to identify peripheral blood biomarkers, predictive of response to immunotherapies against lung cancer, based on whole blood microRNA profiling. Using three well-characterized cohorts consisting of a total of 334 stage IV NSCLC patients, we have defined a 5 microRNA risk score (miRisk) that is predictive of overall survival following immunotherapy in training and independent validation (HR 2.40, 95% CI 1.37–4.19; P < 0.01) cohorts. We have traced the signature to a myeloid origin and performed miRNA target prediction to make a direct mechanistic link to the PD-L1 signaling pathway and PD-L1 itself. The miRisk score offers a potential blood-based companion diagnostic for immunotherapy that outperforms tissue-based PD-L1 staining.

Many applications in the biomedical domain involve the detailed molecular and functional characterization of macro-molecules such as proteins. Where possible, this involves the knowledge of detailed 3D coordinates of every atom within a protein. At the same time, machine learning has become the basis of much innovation within this domain in recent years. There are, however, a few challenges in applying machine learning to 3D protein structures, such as variability in size and high dimensionality of the data. It would therefore be beneficial to be able to map every protein structure to a smaller fixed-dimensional representation that is directly learned from the structure without manual curation. In addition, it would be valuable for biomedical researchers if such approaches would require little method development and instead draw from cutting-edge research such as image classification via deep neural networks. Here, such an approach is outlined that first re-formats protein structures as 2D color images and then applies off-the-shelf neural networks for image classification. It is shown that such neural networks can be trained to effectively encode the CATH protein classification database and that feature vectors extracted from such networks, once trained, can be transferred to a completely new task that is likely to benefit from molecular protein information, namely that of small molecule binding.

Deciphering the effects of nonsynonymous mutations on protein structure is central to many areas of biomedical research and is of fundamental importance to the study of molecular evolution. Much of the investigation of protein evolution has focused on mutations that leave a protein’s folded structure essentially unchanged. However, to evolve novel folds of proteins, mutations that lead to large conformational modifications have to be involved. Unraveling the basic biophysics of such mutations is a challenge to theory, especially when only one or two amino acid substitutions cause a large-scale conformational switch. Among the few such mutational switches identified experimentally, the one between the G_A all-α and G_B α+β folds is extensively characterized; but all-atom simulations using fully transferrable potentials have not been able to account for this striking switching behavior. Here we introduce an explicit-chain model that combines structure-based native biases for multiple alternative structures with a general physical atomic force field, and apply this construct to twelve mutants spanning the sequence variation between G_A and G_B. In agreement with experiment, we observe conformational switching from G_A to G_B upon a single L45Y substitution in the GA98 mutant. In line with the latent evolutionary potential concept, our model shows a gradual sequence-dependent change in fold preference in the mutants before this switch. Our analysis also indicates that a sharp G_A/G_B switch may arise from the orientation dependence of aromatic π-interactions. These findings provide physical insights toward rationalizing, predicting and designing evolutionary conformational switches. The biological functions of globular proteins are intimately related to their folded structures and their associated conformational fluctuations. Evolution of new structures is an important avenue to new functions. Although many mutations do not change the folded state, experiments indicate that a single amino acid substitution can lead to a drastic change in the folded structure. The physics of this switch-like behavior remains to be elucidated. Here we develop a computational model for the relevant physical forces, showing that mutations can lead to new folds by passing through intermediate sequences where the old and new folds occur with varying probabilities. Our approach helps provide a general physical account of conformational switching in evolution and mutational effects on conformational dynamics.

The study of molecular evolution at the level of protein-coding genes often entails comparing large datasets of sequences to infer their evolutionary relationships. Despite the importance of a protein's structure and conformational dynamics to its function and thus its fitness, common phylogenetic methods embody minimal biophysical knowledge of proteins. To underscore the biophysical constraints on natural selection, we survey effects of protein mutations, highlighting the physical basis for marginal stability of natural globular proteins and how requirement for kinetic stability and avoidance of misfolding and misinteractions might have affected protein evolution. The biophysical underpinnings of these effects have been addressed by models with an explicit coarse-grained spatial representation of the polypeptide chain. Sequence–structure mappings based on such models are powerful conceptual tools that rationalize mutational robustness, evolvability, epistasis, promiscuous function performed by ‘hidden’ conformational states, resolution of adaptive conflicts and conformational switches in the evolution from one protein fold to another. Recently, protein biophysics has been applied to derive more accurate evolutionary accounts of sequence data. Methods have also been developed to exploit sequence-based evolutionary information to predict biophysical behaviours of proteins. The success of these approaches demonstrates a deep synergy between the fields of protein biophysics and protein evolution.

Life-history traits controlling the duration and timing of developmental phases in the life cycle jointly determine fitness. Therefore, life-history traits studied in isolation provide an incomplete view on the relevance of life-cycle variation for adaptation. In this study, we examine genetic variation in traits covering the major life history events of the annual species Arabidopsis thaliana: seed dormancy, vegetative growth rate and flowering time. In a sample of 112 genotypes collected throughout the European range of the species, both seed dormancy and flowering time follow a latitudinal gradient independent of the major population structure gradient. This finding confirms previous studies reporting the adaptive evolution of these two traits. Here, however, we further analyze patterns of co-variation among traits. We observe that co-variation between primary dormancy, vegetative growth rate and flowering time also follows a latitudinal cline. At higher latitudes, vegetative growth rate is positively correlated with primary dormancy and negatively with flowering time. In the South, this trend disappears. Patterns of trait co-variation change, presumably because major environmental gradients shift with latitude. This pattern appears unrelated to population structure, suggesting that changes in the coordinated evolution of major life history traits is adaptive. Our data suggest that A. thaliana provides a good model for the evolution of trade-offs and their genetic basis.

Experimental studies have shown that some proteins exist in two alternative native-state conformations. It has been proposed that such bi-stable proteins can potentially function as evolutionary bridges at the interface between two neutral networks of protein sequences that fold uniquely into the two different native conformations. Under adaptive conflict scenarios, bi-stable proteins may be of particular advantage if they simultaneously provide two beneficial biological functions. However, computational models that simulate protein structure evolution do not yet recognize the importance of bi-stability. Here we use a biophysical model to analyze sequence space to identify bi-stable or multi-stable proteins with two or more equally stable native-state structures. The inclusion of such proteins enhances phenotype connectivity between neutral networks in sequence space. Consideration of the sequence space neighborhood of bridge proteins revealed that bi-stability decreases gradually with each mutation that takes the sequence further away from an exactly bi-stable protein. With relaxed selection pressures, we found that bi-stable proteins in our model are highly successful under simulated adaptive conflict. Inspired by these model predictions, we developed a method to identify real proteins in the PDB with bridge-like properties, and have verified a clear bi-stability gradient for a series of mutants studied by Alexander et al. (Proc Nat Acad Sci USA 2009, 106:21149–21154) that connect two sequences that fold uniquely into two different native structures via a bridge-like intermediate mutant sequence. Based on these findings, new testable predictions for future studies on protein bi-stability and evolution are discussed. Proteins are essential molecules for performing a majority of functions in all biological systems. These functions often depend on the three-dimensional structures of proteins. Here, we investigate a fundamental question in molecular evolution: how can proteins acquire new advantageous structures via mutations while not sacrificing their existing structures that are still needed? Some authors have suggested that the same protein may adopt two or more alternative structures, switch between them and thus perform different functions with each of the alternative structures. Intuitively, such a protein could provide an evolutionary compromise between conflicting demands for existing and new protein structures. Yet no theoretical study has systematically tackled the biophysical basis of such compromises during evolutionary processes. Here we devise a model of evolution that specifically recognizes protein molecules that can exist in several different stable structures. Our model demonstrates that proteins can indeed utilize multiple structures to satisfy conflicting evolutionary requirements. In light of these results, we identify data from known protein structures that are consistent with our predictions and suggest novel directions for future investigation.