Abstract Protein aggregation is a critical phenomenon associated with numerous neurodegenerative and systemic diseases. Understanding the propensity of proteins to aggregate is essential for unraveling the molecular basis of these disorders and for the design and engineering of novel proteins, as well as for modulating the activity/stability of enzymatic proteins. Here, we present APR-Score, a novel machine learning–based computational method designed to identify aggregation-prone regions within protein sequences. APR-Score leverages a combination of sequence-based features to predict regions of proteins that are prone to aggregation. It harnessed the information embedded in compiled sequence and structural features to achieve state-of-the-art accuracy. The method was evaluated through rigorous cross-validation on the training dataset and further validated on an independent test dataset. APR-Score demonstrated robustness and reliability in distinguishing aggregation-prone regions from non-aggregating ones, achieving a Matthews correlation coefficient (MCC) of 0.77, precision of 0.89, and F1-Score of 0.87. This tool offers significant value for researchers investigating protein aggregation-related diseases, as it can expedite the identification of aggregation-prone regions, thereby aiding the development of targeted therapies and diagnostic tools. Moreover, computational protein design and engineering efforts can be facilitated by APR-Score–based identification and screening of aggregation-prone protein sequences. APR-Score is available at https://github.com/TII-BRC/APR-Score.
Abstract Human adaptation to extreme cold has been studied predominantly in Arctic populations, leaving the genetic basis of cold resilience in non-Arctic high-latitude regions insufficiently understood. The Amur-Heilong River Basin (AR) in Northeast Asia represents one of the coldest inhabited regions in East Asia, yet its long-term resident populations remain underexplored at the genomic level. Here, we compiled high-coverage whole-genome sequences for 78 individuals representing three indigenous AR populations (Daur, Evenki, and Oroqen) to investigate their demographic history and genetic adaptation to cold environments. Population genomic analyses revealed distinct ancestral compositions, with Evenki and Oroqen showing closer genetic ties to Siberian populations, while Daur exhibited greater admixture with northern East Asians. Demographic reconstructions identified shared and prolonged bottlenecks across AR populations between ∼ 10,000 and 3000 years ago. Genome-wide selection scans uncovered candidate loci associated with cold adaptation, including SORBS2, PRKAG2, KCNMA1, PRKG1, NTM, CHST11, OCA2, RYR3, OTOA, and PKIG, many showing strong correlations between allele frequency and cold-related climatic variables. Notably, a Neanderthal-derived allele at CHST11 (rs11112181) likely contributed to cold-adaptive variation. Functional enrichment analyses suggest that cold adaptation in AR populations may preferentially involve neuro-sensory regulation and tissue integrity, rather than the classical metabolic thermogenesis pathways emphasized in Arctic populations. Integrating population history, selection signals, and archaic introgression, our results indicate that cold resilience in the AR reflects a composite of ancestral adaptation, demographic processes, and limited archaic contributions. These findings expand current understanding of human cold adaptation beyond the Arctic and highlight alternative evolutionary strategies in extreme environments.
Abstract While the systemic lupus erythematosus (SLE), a severe autoimmune disease, is known to relapse frequently after treatment, none of the previous single-cell studies have examined the immune dynamics and their potential contribution to this relapse in the real clinical setting. To address this issue, here we employed single-cell multi-omics profiling to delineate immune repertoire dynamics during the immunosuppressant-based standard of care. Further analysis showed that the canonical therapy fails to restore various critical abnormalities, including the drop in mutational load of B and plasma cells, and the clonal expansion of CD4+ Th1 and CD8+ memory T cells. We also integrated our data with published SLE datasets to compile a comprehensive SLE immune atlas of about 1.6 million cells. Our study provides valuable data and sheds light on the underlying multifaceted unresolved dysregulation upon SLE immunosuppressant therapy under the real clinical setting; such dysregulation may further contribute to the high post-treatment relapse rate and thus call for additional target-specific treatment strategies.
Abstract Understanding the genetic regulation of plasma inflammatory proteins is essential for bridging the gap between genetics and disease biology. While studies in European populations have highlighted significant genetic influences on inflammatory protein levels, African populations remain underrepresented in immunogenetic research. Here, we compare the genetic determinants of 72 inflammatory proteins quantified using the Olink Target 96 Inflammation panel in two cohorts from the Human Functional Genomics Project. The study included 286 healthy Tanzanians (146 males and 140 females) and 340 Dutch participants (175 males and 165 females). Genotyping was conducted using the Illumina Global Screening Array (GSA) for the Tanzanian cohort and the HumanOmniExpressExome-8 for the Dutch cohort. Our genome-wide protein quantitative trait locus (pQTL) analysis identified 10 genome-wide significant FDR-corrected pQTLs in each population, including a missense variant affecting cluster of differentiation 6 (CD6) concentrations, which is implicated in multiple autoimmune and inflammatory diseases. We identified three trans- and seven cis-pQTLs in the Tanzanian cohort, while three cis- and seven trans-pQTLs were identified in the Dutch cohort. However, we observed minimal overlap (< 2%) in pQTLs between the two populations, with colocalization analysis confirming only nine shared loci. Additionally, we replicated several pQTLs identified in the UK-BIOBANK study across African and European ancestries. These findings underscore the genetic diversity underlying inflammatory protein regulation and highlight the importance of population-specific studies to better understand immune function, disease susceptibility, and tailored health interventions across diverse ancestries.
Accurate prediction of peptide binding to human leukocyte antigen class I (HLA-I) molecules is critical for advancing immunological research, particularly in vaccine design and immunotherapy. However, limitations in model performance, interpretability, and dataset quality impede the widespread adoption of existing predictive tools. Here, we present a comprehensive evaluation of 17 HLA-I peptide binding prediction models, utilizing a meticulously curated dataset comprising over 290,000 peptides spanning 44 HLA-I alleles. We assessed model accuracy, robustness, and interpretability, employing explainability techniques such as SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) to elucidate underlying prediction mechanisms. Our results reveal substantial performance disparities, with self-attention-based models, including STMHCpan and BigMHC, exhibiting superior accuracy. Notably, the capsule network model CapsNet-MHC_AN demonstrated robust performance. Models trained on eluted ligand datasets outperformed those relying on binding affinity data, underscoring the critical role of high-quality training data. Ensemble and multi-algorithm approaches further improved prediction reliability. These findings highlight the need for ongoing innovation in model architecture, integration of diverse and high-quality datasets, and incorporation of structural predictors to develop more accurate, interpretable, and clinically applicable HLA-I peptide binding prediction tools.
