Xikang Feng 冯喜康

Abstract Recurrent copy number alterations (CNAs) are fundamental drivers of tumorigenesis, yet identifying them reliably remains a challenge due to the extreme variability in their genomic scale and context. Current methods often struggle to balance sensitivity across focal, segmental, and arm-level events. Here, we present CRESCENT, a deep learning framework designed to detect recurrent CNAs by integrating multi-scale sampling with convolutional neural networks and self-attention mechanisms. By processing copy number profiles from 7689 cases across 20 The Cancer Genome Atlas (TCGA) cancer projects, CRESCENT learns to distinguish recurrent drivers from background noise through parallel feature fusion. In rigorous leave-one-project-out cross-validation, the model demonstrated robust generalization, achieving area under the curves of 0.894–0.967 for amplifications and 0.804–0.929 for deletions in representative cohorts (Bladder Urothelial Carcinoma, Sarcoma, Glioblastoma Multiforme, Uterine Corpus Endometrial Carcinoma). Finally, extending beyond the TCGA-specific cross-validation, we trained a unified pan-cancer model to assess CRESCENT’s generalizability on simulated datasets and independent, non-TCGA cancer cohorts (CGCI and TARGET). Benchmarking against standard tools, including GISTIC2 and RUBIC, reveals that CRESCENT offers superior detection balance, identifying the highest total number of significant events across focal and broad scales. Moreover, extensive focal gene expression validation and pathway annotation, coupled with survival analysis, highlight that CRESCENT identifies critical oncogenic drivers and prognostic markers that conventional statistical methods often overlook. In all, CRESCENT provides a highly sensitive, generalized approach for decoding tumor evolution.

Abstract Copy number aberrations (CNAs) are critical drivers of genomic diversity in oncology, where recurrent CNAs frequently underlie tumorigenesis. However, existing public resources are limited in their somatic CNA specificity, breadth across multiple data modalities, and support for recurrent CNAs with online functional annotation and interactive visualization. Here, we present CNAScope (https://cna.fengslab.com/), a database that curates and functionally annotates over 3 954 361 CNA profiles and 3 946 319 metadata from 810 datasets, 174 464 samples, 3 018 672 single cells, and 764 232 spatial cells/spots, spanning 77 cancer subtypes from eight data sources and 55 cancer initiatives and institutions. CNAScope offers downloadable CNA annotations and interactive visualizations at bin, gene, and pathway term levels, including phylogenetic inference, clustering, dimension reduction, and focal/consensus CNA detection. Users can explore data through interactive heatmaps, phylogenetic trees, embedding plots, CN charts, and focal/consensus plots, or upload and annotate their own CNAs in real time. In all, with its large curated data volume and rich annotation capabilities, CNAScope serves as a vital resource for accelerating cancer research.

Abstract Microorganisms, including bacteria, archaea, fungi, and viruses, are the most taxonomically diverse and ecologically dominant life forms on Earth, playing critical roles in ecosystems, human health, and industrial applications. While existing microbial databases such as BV-BRC and IMG archive both monoisolate and metagenome-assembled genomes (MAGs) across domains, challenges remain in standardized, multi-level annotations and interactive tools for all microbial groups. Here, we present MicrobialScope (https://microbial.deepomics.org/), a comprehensive microbial genomic platform that integrates large-scale genome collections, multilevel annotations, and interactive visualizations. MicrobialScope harbors 2 411 503 bacterial, 24 472 archaeal, 20 203 fungal, and 188 267 viral genomes derived from both monoisolate assemblies and MAGs. Integrating 15 state-of-the-art bioinformatics tools and 10 specialized databases, MicrobialScope provides extensive annotations encompassing basic genomic features, genomic element prediction (e.g., genes, tRNAs, tmRNAs, CRISPR–Cas and anti-CRISPR elements, secondary metabolite biosynthetic clusters, signal peptides, and transmembrane proteins), and functional and structural annotations. This includes 1 072 114 935 proteins with diverse annotations, 24 640 186 tRNAs and tmRNAs, 140 888 CRISPR–Cas systems, 173 256 anti-CRISPR elements, 105 121 secondary metabolite biosynthetic clusters, 13 235 096 signal peptides, and 50 811 729 transmembrane proteins. In addition, MicrobialScope offers unrestricted access to all data resources, interactive visualization tools, and built-in online analytical modules for intuitive exploration and comparative analysis. With its extensive genome collection, comprehensive annotations, and user-friendly interface, MicrobialScope serves as a scalable platform to advance genome research across diverse microbial domains.

Abstract The complexity of T cell receptor (TCR) sequences, particularly within the complementarity-determining region 3 (CDR3), requires efficient embedding methods for applying machine learning to immunology. While various TCR CDR3 embedding strategies have been proposed, the absence of their systematic evaluations created perplexity in the community. Here, we extracted CDR3 embedding models from 19 existing methods and benchmarked these models with four curated datasets by accessing their impact on the performance of TCR downstream tasks, including TCR-epitope binding affinity prediction, epitope-specific TCR identification, TCR clustering, and visualization analysis. We assessed these models utilizing eight downstream classifiers and five downstream clustering methods, with the performance measured by a diverse range of metrics for precision, robustness, and usability. Overall, handcrafted embeddings outperformed data-driven ones in modeling TCR-epitope interactions. To further refine our comparative findings, we developed an all-in-one TCR CDR3 embedding package comprising all evaluated embedding models. This package will assist users in easily selecting suitable embedding models for their data.