MOTIVATION: B-cell receptors (BCRs) and gene expression profiles are two distinct yet complementary modalities of B cells. However, most analyses treat them independently. Here, we present CoMBCR, a B-cell embedding tool that co-learns BCRs and gene expressions, representing data within a unified latent space for downstream analysis. RESULTS: We applied CoMBCR to 126,791 B cells from diverse datasets with matched BCRs and gene expressions. First, CoMBCR outperforms the methods solely encoding BCRs in capturing B-cell biological features, achieving at least 0.1 improvement in Matthews Correlation Coefficient on a SARS-CoV-2 binding prediction task. Second, CoMBCR reveals active immune responses and CDR3 motif preferences through modality gap analysis in SARS-CoV-2-specific memory B cells. Moreover, when supported by spatial transcriptomics data, CoMBCR accurately traces the developmental trajectories of malignant B cells and uncovers transcriptional patterns associated with their survival within lymphoma patients. AVAILABILITY AND IMPLEMENTATION: The CoMBCR software is publicly available under the MIT License at https://github.com/deepomicslab/CoMBCR.git. CONTACT: [email protected].

MOTIVATION: B-cell lineage trees describe the evolutionary process of immunoglobulin genes during affinity maturation. Existing methods for building B-cell lineage trees generally do not guarantee the parent-to-child inheritance and accumulation of advantageous mutations under successive rounds of somatic hypermutation (SHM) and selection, and are often incompatible with repertoire input. RESULTS: To address previous limitations, we developed AffMB (Affinity Maturation of B-cell receptor), a comprehensive toolkit for tracking affinity maturation through the generation and visualization of SHM-ordered, inheritance-based B-cell lineage trees from single-cell or bulk B-cell receptor sequencing data. The SHM-ordered inheritance tree algorithm outperformed state-of-the-art benchmarks in simulations. When applied to single-cell data from BNT162b2 vaccination (n = 42), AffMB demonstrated the ability to infer immunization responses and showed the feasibility of identifying potential high-affinity antibody sequences. AVAILABILITY AND IMPLEMENTATION: AffMB is an open-source Python package that supports contig FASTA or AIRR rearrangement TSV inputs. The source code for AffMB is freely available at https://github.com/deepomicslab/AffMB.

Machine learning (ML) is increasingly being used to guide biological discovery in biomedicine such as prioritizing promising small molecules in drug discovery. In those applications, ML models are used to predict the properties of biological systems, and researchers use these predictions to prioritize candidates as new biological hypotheses for downstream experimental validations. However, when applied to unseen situations, these models can be overconfident and produce a large number of false positives. One solution to address this issue is to quantify the model's prediction uncertainty and provide a set of hypotheses with a controlled false discovery rate (FDR) pre-specified by researchers. We propose CPEC, an ML framework for FDR-controlled biological discovery. We demonstrate its effectiveness using enzyme function annotation as a case study, simulating the discovery process of identifying the functions of less-characterized enzymes. CPEC integrates a deep learning model with a statistical tool known as conformal prediction, providing accurate and FDR-controlled function predictions for a given protein enzyme. Conformal prediction provides rigorous statistical guarantees to the predictive model and ensures that the expected FDR will not exceed a user-specified level with high probability. Evaluation experiments show that CPEC achieves reliable FDR control, better or comparable prediction performance at a lower FDR than existing methods, and accurate predictions for enzymes under-represented in the training data. We expect CPEC to be a useful tool for biological discovery applications where a high yield rate in validation experiments is desired but the experimental budget is limited.

Deep neural networks (DNNs) have been shown to be vulnerable to adversarial examples. Moreover, the transferability of the adversarial examples has received broad attention in recent years, which means that adversarial examples crafted by a surrogate model can also attack unknown models. This phenomenon gave birth to the transfer-based adversarial attacks, which aim to improve the transferability of the generated adversarial examples. In this paper, we propose to improve the transferability of adversarial examples in the transfer-based attack via masking unimportant parameters (MUP). The key idea in MUP is to refine the pretrained surrogate models to boost the transfer-based attack. Based on this idea, a Taylor expansion-based metric is used to evaluate the parameter importance score and the unimportant parameters are masked during the generation of adversarial examples. This process is simple, yet can be naturally combined with various existing gradient-based optimizers for generating adversarial examples, thus further improving the transferability of the generated adversarial examples. Extensive experiments are conducted to validate the effectiveness of the proposed MUP-based methods.

The specificity of a T-cell receptor (TCR) repertoire determines personalized immune capacity. Existing methods have modeled the qualitative aspects of TCR specificity, while the quantitative aspects remained unaddressed. We developed a package, TCRanno, to quantify the specificity of TCR repertoires. We created deep-learning-based, epitope-aware vector embeddings to infer individual TCR specificity. Then we aggregated clonotype frequencies of TCRs to obtain a quantitative profile of repertoire specificity at epitope, antigen and organism levels. Applying TCRanno to 4195 TCR repertoires revealed quantitative changes in repertoire specificity upon infections, autoimmunity and cancers. Specifically, TCRanno found cytomegalovirus-specific TCRs in seronegative healthy individuals, supporting the possibility of abortive infections. TCRanno discovered age-accumulated fraction of severe acute respiratory syndrome coronavirus 2 specific TCRs in pre-pandemic samples, which may explain the aggressive symptoms and age-related severity of coronavirus disease 2019. TCRanno also identified the encounter of Hepatitis B antigens as a potential trigger of systemic lupus erythematosus. TCRanno annotations showed capability in distinguishing TCR repertoires of healthy and cancers including melanoma, lung and breast cancers. TCRanno also demonstrated usefulness to single-cell TCRseq+gene expression data analyses by isolating T-cells with the specificity of interest.

Mobile Edge Computing (MEC) is a promising solution to tackle the upcoming computing tsunami in 5G era, by effectively utilizing the idle resource at the mobile edge. In this work, we study such an MEC scenario, where mobile devices at edge share their heterogeneous resources with each other, hence forming a multi-dimensional resource crowdsourcing (sharing) framework. We are interested in the problem of how to optimally offload tasks to mobile devices under this framework, aiming at minimizing the total energy cost and maximizing the overall task completion. To study the problem, we first propose a general task model, where each task is divided into multiple sequential subtasks according to their functionalities as well as resource requirements. Then, based on the task model, we propose a Joint Energy Consumption and Task Failure Probability Minimization Problem, which decides when and where each subtask will be offloaded to. The problem is challenging to solve, mainly due to the inherent constraints between the scheduling of different subtasks. Therefore, we propose several linearization methods to relax the constraints, and convert the original problem into an integer linear programming (ILP), which can be solved by many classic methods effectively. We further perform simulations, which show that our proposed solution outperforms the existing solutions (with indivisible tasks or without resource sharing) in terms of both the total cost and the task failure probability. Precisely, our proposed solution can reduce the total cost by 25%~85% and the task failure probability by 10%~35%.

Deep learning has been widely applied in the field of medical image processing. However, compared with flourishing visual tasks in natural images, the progress achieved in pathological images is not remarkable, and detection and segmentation, which are among basic tasks of computer vision, are regarded as two independent tasks. In this paper, we make full use of existing datasets and construct a triplet-path network using dilated convolutions to cooperatively accomplish one-stage object detection and nuclei segmentation for general pathological images. First, in order to meet the requirement of detection and segmentation, a novel structure called triplet feature generation (TFG) is designed to extract high-resolution and multiscale features, where features from different layers can be properly integrated. Second, considering that pathological datasets are usually small, a location-aware and partially truncated loss function is proposed to improve the classification accuracy of datasets with few images and widely varying targets. We compare the performance of both object detection and instance segmentation with state-of-the-art methods. Experimental results demonstrate the effectiveness and efficiency of the proposed network on two datasets collected from multiple organs.

MOTIVATION: A variety of in silico tools have been developed and frequently used to aid high-throughput rapid variant classification, but their performances vary, and their ability to classify variants of uncertain significance were not systemically assessed previously due to lack of validation data. This has been changed recently by advances of functional assays, where functional impact of genetic changes can be measured in single-nucleotide resolution using saturation genome editing (SGE) assay. RESULTS: We demonstrated the neural network model AIVAR (Artificial Intelligent VARiant classifier) was highly comparable to human experts on multiple verified datasets. Although highly accurate on known variants, AIVAR together with CADD and PhyloP showed non-significant concordance with SGE function scores. Moreover, our results indicated that neural network model trained from functional assay data may not produce accurate prediction on known variants. AVAILABILITY AND IMPLEMENTATION: All source code of AIVAR is deposited and freely available at https://github.com/TopGene/AIvar. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.