Predicting disease risk from DNA presents an unprecedented emerging challenge as biobanks approach population scale sizes ($N>10^6$ individuals) with ultra-high-dimensional features ($L>10^5$ genotypes). Current methods, often linear and reliant on summary statistics, fail to capture complex genetic interactions and discard valuable individual-level information. We introduce **PRSformer**, a scalable deep learning architecture designed for end-to-end, multitask disease prediction directly from million-scale individual genotypes. PRSformer employs neighborhood attention, achieving linear $O(L)$ complexity per layer, making Transformers tractable for genome-scale inputs. Crucially, PRSformer utilizes a stacking of these efficient attention layers, progressively increasing the effective receptive field to model local dependencies (e.g., within linkage disequilibrium blocks) before integrating information across wider genomic regions. This design, tailored for genomics, allows PRSformer to learn complex, potentially non-linear and long-range interactions directly from raw genotypes. We demonstrate PRSformer's effectiveness using a unique large private cohort ($N \approx 5$M) for predicting 18 autoimmune and inflammatory conditions using $L \approx 140$k variants. PRSformer significantly outperforms highly optimized linear models trained on the *same individual-level data* and state-of-the-art summary-statistic-based methods (LDPred2) derived from the *same cohort*, quantifying the benefits of non-linear modeling and multitask learning at scale. Furthermore, experiments reveal that the advantage of non-linearity emerges primarily at large sample sizes ($N > 1$M), and that a multi-ancestry trained model improves generalization, establishing PRSformer as a new framework for deep learning in population-scale genomics.

Instance-level image classification tasks have traditionally relied on single-instance labels to train models, e.g., few-shot learning and transfer learning. However, set-level coarse-grained labels that capture relationships among instances can provide richer information in real-world scenarios. In this paper, we present a novel approach to enhance instance-level image classification by leveraging set-level labels. We provide a theoretical analysis of the proposed method, including recognition conditions for fast excess risk rate, shedding light on the theoretical foundations of our approach. We conducted experiments on two distinct categories of datasets: natural image datasets and histopathology image datasets. Our experimental results demonstrate the effectiveness of our approach, showcasing improved classification performance compared to traditional single-instance label-based methods. Notably, our algorithm achieves 13\% improvement in classification accuracy compared to the strongest baseline on the histopathology image classification benchmarks. Importantly, our experimental findings align with the theoretical analysis, reinforcing the robustness and reliability of our proposed method. This work bridges the gap between instance-level and set-level image classification, offering a promising avenue for advancing the capabilities of image classification models with set-level coarse-grained labels.

Proximity sequencing (Prox-seq) simultaneously measures gene expression, protein expression and protein complexes on single cells. Using information from dual-antibody binding events, Prox-seq infers surface protein dimers at the single-cell level. Prox-seq provides multi-dimensional phenotyping of single cells in high throughput, and was recently used to track the formation of receptor complexes during cell signaling and discovered a novel interaction between CD9 and CD8 in naïve T cells. The distribution of protein abundance can affect identification of protein complexes in a complicated manner in dual-binding assays like Prox-seq. These effects are difficult to explore with experiments, yet important for accurate quantification of protein complexes. Here, we introduce a physical model of Prox-seq and computationally evaluate several different methods for reducing background noise when quantifying protein complexes. Furthermore, we developed an improved method for analysis of Prox-seq data, which resulted in more accurate and robust quantification of protein complexes. Finally, our Prox-seq model offers a simple way to investigate the behavior of Prox-seq data under various biological conditions and guide users toward selecting the best analysis method for their data.

Reproducibility of results obtained using ribonucleic acid (RNA) data across labs remains a major hurdle in cancer research. Often, molecular predictors trained on one dataset cannot be applied to another due to differences in RNA library preparation and quantification, which inhibits the validation of predictors across labs. While current RNA correction algorithms reduce these differences, they require simultaneous access to patient-level data from all datasets, which necessitates the sharing of training data for predictors when sharing predictors. Here, we describe SpinAdapt, an unsupervised RNA correction algorithm that enables the transfer of molecular models without requiring access to patient-level data. It computes data corrections only via aggregate statistics of each dataset, thereby maintaining patient data privacy. Despite an inherent trade-off between privacy and performance, SpinAdapt outperforms current correction methods, like Seurat and ComBat, on publicly available cancer studies, including TCGA and ICGC. Furthermore, SpinAdapt can correct new samples, thereby enabling unbiased evaluation on validation cohorts. We expect this novel correction paradigm to enhance research reproducibility and to preserve patient privacy.

Single cell RNA sequencing (scRNAseq) can be used to infer a temporal ordering of cellular states. Current methods for the inference of cellular trajectories rely on unbiased dimensionality reduction techniques. However, such biologically agnostic ordering can prove difficult for modeling complex developmental or differentiation processes. The cellular heterogeneity of dynamic biological compartments can result in sparse sampling of key intermediate cell states. To overcome these limitations, we develop a supervised machine learning framework, called Pseudocell Tracer, which infers trajectories in pseudospace rather than in pseudotime. The method uses a supervised encoder, trained with adjacent biological information, to project scRNAseq data into a low-dimensional manifold that maps the transcriptional states a cell can occupy. Then a generative adversarial network (GAN) is used to simulate pesudocells at regular intervals along a virtual cell-state axis. We demonstrate the utility of Pseudocell Tracer by modeling B cells undergoing immunoglobulin class switch recombination (CSR) during a prototypic antigen-induced antibody response. Our results revealed an ordering of key transcription factors regulating CSR to the IgG1 isotype, including the concomitant expression of Nfkb1 and Stat6 prior to the upregulation of Bach2 expression. Furthermore, the expression dynamics of genes encoding cytokine receptors suggest a poised IL-4 signaling state that preceeds CSR to the IgG1 isotype.

We bring together into a common framework three key ideas - multi-scale medical image analysis, the attention mechanism, and hyperbolic embeddings. The formulation and evaluation of hyperbolic-attention models for multi-scale medical image analysis have not been previously explored. In this paper, we evaluate a hyperbolic-attention model on two classification tasks using histopathology image datasets. The experiments show improvement compared to other commonly used models. Our method directly captures the multi-scale structure of histopathology images, and we speculate that the hyperbolic attention mechanism naturally singles out one or more structures at one or more scales that are most discriminatory.

Motivation: High-throughput experimental techniques have produced a large amount of protein-protein interaction (PPI) data, but their coverage is still low and the PPI data is also very noisy. Computational prediction of PPIs can be used to discover new PPIs and identify errors in the experimental PPI data. Results: We present a novel deep learning framework, DPPI, to model and predict PPIs from sequence information alone. Our model efficiently applies a deep, Siamese-like convolutional neural network combined with random projection and data augmentation to predict PPIs, leveraging existing high-quality experimental PPI data and evolutionary information of a protein pair under prediction. Our experimental results show that DPPI outperforms the state-of-the-art methods on several benchmarks in terms of area under precision-recall curve (auPR), and computationally is more efficient. We also show that DPPI is able to predict homodimeric interactions where other methods fail to work accurately, and the effectiveness of DPPI in specific applications such as predicting cytokine-receptor binding affinities. Availability and implementation: Predicting protein-protein interactions through sequence-based deep learning): https://github.com/hashemifar/DPPI/. Supplementary information: Supplementary data are available at Bioinformatics online.

Motivation: The B-cell receptor enables individual B cells to identify diverse antigens, including bacterial and viral proteins. While advances in RNA-sequencing (RNA-seq) have enabled high throughput profiling of transcript expression in single cells, the unique task of assembling the full-length heavy and light chain sequences from single cell RNA-seq (scRNA-seq) in B cells has been largely unstudied. Results: We developed a new software tool, BASIC, which allows investigators to use scRNA-seq for assembling BCR sequences at single-cell resolution. To demonstrate the utility of our software, we subjected nearly 200 single human B cells to scRNA-seq, assembled the full-length heavy and the light chains, and experimentally confirmed these results by using single-cell primer-based nested PCRs and Sanger sequencing. Availability and Implementation: http://ttic.uchicago.edu/∼aakhan/BASIC Contact: [email protected] Supplementary Information: Supplementary data are available at Bioinformatics online.