Papers for

biomedical data engineers

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

CytoWave improves prediction of cell gene responses to changes

CytoWave: Perturbation-Centric Pretraining for Single-Cell Response Prediction

Abstract: Predicting gene expression under perturbations from single-cell data is a fundamental task for modeling cellular responses, with applications in discovering gene function and designing therapeutics. However, existing methods often struggle to generalize across datasets due to differences in experimental protocols, biological contexts, and species. We present \textbf{CytoWave}, a perturbation-centric pretraining framework for single-cell genetic perturbation response prediction. We leverage large-scale cross-species perturbation datasets unified via orthology and train CytoWave in a two-stage manner. In the first stage, we learn reference control cell states via masked gene expression recovery on unperturbed cells. In the second stage, we pretrain the model to predict perturbed gene expression given control states and perturbation signals, together with a contrastive objective that aligns latent representations of perturbation responses with perturbation signals. We perform strict dataset-held-out evaluation across multiple condition combinations and case studies of pathway activity changes, demonstrating consistent and effective performance of CytoWave in predicting single-cell genetic perturbation responses.

Mon 2 NovMachine Learning
The gist
Predicting how individual cells change their gene activity when affected by different experiments or treatments is very important but hard because data can vary widely. The authors created CytoWave, a method that learns from large collections of gene data across species to better predict these changes. They train it in two parts: first by learning normal cell states, then by learning how cells respond to specific changes. Tests show CytoWave predicts changes well even on new types of data it did not see before.
Open → 2611.02311v1

Biology guided method improves gene expression prediction from tissue images

Biology-Guided Context-Aware Residual Propagation for Spatial Gene Expression Prediction from Histology

Abstract: Spatial transcriptomics provides spatially resolved molecular profiling, but its cost, limited throughput and experimental complexity hinder routine large-scale use. Predicting spatial gene expression from hematoxylin-and-eosin-stained histology images offers a scalable alternative, yet remains challenging because histology provides only an indirect morphological view of molecular activity. Existing methods typically decode expression from visual features in a single stage, even when enhanced with spatial aggregation, retrieval or multi-resolution fusion. This formulation is restrictive because spot-level expression reflects not only local morphology but also cellular identity, neighborhood composition and tissue-level organization. We propose BCARP, a Biology-guided Context-Aware Residual Propagation framework for histology-guided spatial transcriptomics prediction. Instead of directly predicting gene expression from histology embeddings, BCARP treats them as morphology visual priors and progressively refines spot representations in a transcriptomics-oriented latent space. Each refinement step integrates nuclei-guided local cellular context with prototype-guided global context propagation, enabling information exchange across both neighboring and spatially distant spots. A pattern-scale decoder further separates spatial expression into pattern prediction from raw expression with scale calibration. Evaluated on three ST datasets and ten HEST-1k cohorts, BCARP achieves the highest mean correlation-based performance against other baselines. Numerical results support biology-guided local-global residual propagation as an effective alternative to visual decoding for histology-to-transcriptomics prediction.

Mon 26 OctMachine Learning
The gist
Spatial transcriptomics shows where genes are active in a tissue but is expensive and complex to do on a large scale. The authors propose a method called BCARP that predicts these gene activity patterns using regular stained tissue images. Instead of guessing gene activity directly from images, BCARP improves its predictions step-by-step by considering cell types nearby and overall tissue structure. This approach worked better than others when tested on several datasets.
Open → 2610.35167v1

Boruta gene selection improves with uncapped tree depth and quality filtering

Tree Depth and Sample-Quality-Associated Features: Two Hidden Factors in Boruta Selection for RNA-Seq Data Classification

Abstract: Boruta is a widely used method for selecting gene panels from RNA sequencing data, and its most used implementation recommends Random Forest trees of depth 3 to 7, an untested setting. We tested whether this setting changes which genes are selected, using pan-cancer data from The Cancer Genome Atlas (10,469 tumors, 33 cancer types) and breast cancer PAM50 subtyping (981 tumors, 5 subtypes). A leakage-free protocol fitted every label-dependent step inside training folds, measured every effect against seed-to-seed variation, and held out a locked test set per task. In pan-cancer data, shallow trees changed the selection far more than a new seed did: the top-500 overlap with unlimited-depth selections was 0.37 at depth 5, against 0.84 to 0.93 between seeds at depth 5 or unlimited depth. Shallow trees confirmed 3.5 times fewer genes, missed markers of large classes and produced less accurate 50-gene panels (balanced accuracy 0.809 vs 0.889). The effect vanished once the cap reached about 60 to 75% of the depth that unrestricted trees grow to, set mainly by the number of classes, not sample size, and leaving depth uncapped cost no accuracy in either task. The effect replicated in a second cross-validation repeat and under permutation importance. About a third of pre-filtered candidates also tracked sample quality within class; a within-class filter removed them without loss of test accuracy (balanced accuracy 0.943 vs 0.942 with 1,000 genes). We recommend uncapped trees, or a cap chosen from the realized tree depth, and a within-class quality filter before Boruta.

Fri 23 OctMachine Learning
The gist
Selecting important genes from RNA sequencing data helps understand diseases like cancer. The authors found that limiting the depth of decision trees in the popular Boruta method changes which genes are picked and lowers accuracy. They showed that allowing deeper trees and filtering genes related to sample quality improves the gene panel selection without losing accuracy. This helps build better tools for classifying cancer types based on gene expression.
Open → 2610.33307v1

Medical image clustering improved with visual and topological refinement

VTRL-Med: Progressive Visual Refinement with Topology-Guided Regularization for Medical Image Clustering

Abstract: Deep clustering offers a promising approach for analyzing large unlabeled medical image datasets, yet prevailing methods still face two limitations: they may neglect class-agnostic structural priors that capture object boundaries and regions, and they do not explicitly encode topological patterns (e.g., connectivity and holes). These sources of information play distinct and complementary roles: visual-structural features provide the primary discriminative representation for clustering, while topological features provide additional global geometric cues for residual visual ambiguities. We introduce VTRL-Med, a progressive visual-structural learning framework with topology-guided regularization for medical image clustering. VTRL-Med first adapts the Segment Anything Model (SAM) to extract image features that retain structural cues inherited from mask-supervised pretraining. It then generates reliable pseudo-labels through a two-step filtering process that combines prototype-based selection with k-nearest neighbor consistency validation. Finally, it progressively refines the visual branch while an auxiliary topology branch, constructed from persistent homology (PH) descriptors, provides additional geometric regularization during training. Across 13 medical image benchmarks, VTRL-Med achieves state-of-the-art clustering performance, improving average accuracy by 4.48% over existing methods. On three representative datasets, topology-guided regularization further improves accuracy by 1.3--4.3% over visual refinement alone.

Fri 23 OctComputer Vision and Pattern Recognition
The gist
Clustering large collections of medical images into meaningful groups is challenging because many images lack labels and have complex structures. The authors introduce a method called VTRL-Med that first extracts detailed visual features showing object shapes, then uses geometric information about image topology to guide learning. This helps the method understand both local details and overall shape patterns in the images, leading to better clustering results. Tests on many medical imaging datasets show this approach improves accuracy compared to existing methods.
Open → 2610.32884v1

Automated patient-specific phenotype detection from full biomedical articles

Overview of CHIP 2026 Shared Task 1: Patient-Specific Phenotype Concept Recognition in Full-Text Biomedical Literature

Abstract: Automatic phenotype information extraction plays an important role in rare disease diagnosis, genotype-phenotype analysis, and precision medicine. However, accurately identifying phenotype concepts and determining their associations with specific patients remain challenging, particularly in full-text biomedical articles involving multiple patients and complex clinical descriptions. This paper presents an overview of CHIP 2026 Shared Task 1, which aims to promote the development of methods for patient-specific phenotype concept recognition from full-text biomedical literature. The task consists of two subtasks: full-text phenotype concept recognition and patient-specific phenotype association. For this evaluation, we used 200 annotated full-text biomedical articles from the PatientPheX dataset, containing 506 patients, 15,193 phenotype mentions, and 5,656 patient-phenotype associations. A total of 455 teams registered for the competition, with 189 and 103 valid submissions received on the Phase A and Phase B leaderboards, respectively. The best-performing system achieved an overall score of 0.7389 on the final test set, outperforming the baseline by 8.37 percentage points. The top-performing teams explored diverse approaches combining biomedical language models, large language models (LLMs), and ensemble strategies. The evaluation results highlight both the potential of these approaches and the remaining challenges in patient attribution and phenotype normalization from full-text biomedical literature. The post-evaluation phase remains open at https://tianchi.aliyun.com/competition/entrance/532512.

Wed 21 OctComputation and LanguageArtificial Intelligence
The gist
Understanding detailed patient information in medical articles is hard because many patients and complex descriptions appear together. The authors organized a challenge called CHIP 2026 to improve how computers find and link phenotype information to specific patients in full biomedical texts. They used a dataset with many articles, patients, and noted phenotype details. Many teams joined and tried different approaches using advanced language models, and the best systems were notably better than the baseline. The results show progress but also highlight there are still challenges in accurately linking phenotypes to the right patients in long medical texts.
Open → 2610.29606v1

Biomedical retrieval augmented generation improves question answering efficiency

Large-Scale Biomedical RAG: Corpus Scaling, Document Representation, and Efficiency Trade-offs

Abstract: We present a controlled, large-scale evaluation of how corpus scale, document representation, and reranking affect the effectiveness and efficiency of biomedical Retrieval-Augmented Generation (RAG) systems. Retrieval collections were progressively expanded while retaining the relevant source documents, enabling consistent comparisons across corpus sizes. We compared abstract, flat full-text, and hierarchical full-text representations across first-stage retrieval, reranking, and end-to-end multiple-choice question answering. Configurations selected on development data were evaluated on five external biomedical benchmarks and a held-out test set, while retrieval latency and index storage were measured to quantify operational cost. Performance patterns observed with small or controlled collections did not consistently persist as collections grew, and the usefulness of reranking depended on the corpus and retrieval setting. RAG improved three of four model configurations across the external benchmarks, with mean gains of up to 9.6 percentage points, and improved all four configurations on the held-out test set by 10.2-30.3 points. Abstract retrieval performed best when reference evidence was abstract-aligned, whereas full-text representations were generally more effective when richer evidence was needed. Hierarchical representation matched or outperformed flat full-text retrieval in most end-to-end comparisons while requiring 67% fewer indexed units and 44-56% less storage. These findings show that corpus scale should be treated as an explicit evaluation variable in biomedical RAG, and that the most appropriate design depends on evidence format, model behavior, and operational constraints.

Mon 19 OctComputation and LanguageArtificial IntelligenceInformation Retrieval
The gist
Finding information quickly from lots of medical documents is tricky because of how much content there is and how it's organized. The authors looked at different ways to search through medical texts and tested how well each worked for answering medical questions. They found that retrieving just the summaries is best when the answers are concise, while using the full documents helps when more detailed information is needed. They also showed that organizing the full texts hierarchically saves storage space and speeds up searching without losing accuracy. Their work helps show how the size and format of medical collections affect the best search approach.
Open → 2610.26339v1

RAGenome improves gene finding by modeling long genome sequences

RAGenome: Scaling Retrieval-Based Genomic Language Models to Long Contexts

Abstract: The genome holds the blueprint that governs the biological properties of the cell. Consequently, advancing our knowledge of genomic function is crucial both for a broader understanding of biology and for continued biomedical advances. The success of large language models on natural language and protein sequences has motivated similar efforts on genomic data. However, standard genomic language models (gLMs) often require extremely large computational resources and still fall behind traditional methods on some downstream tasks. Recently, MSA-based pretraining has been proposed as an efficient alternative, but existing models are limited to short input contexts, restricting their use to short-range tasks, such as variant effect prediction. In this work, we present RAGenome, the first retrieval-based gLM that scales pretraining to longer contexts (100$\times$ longer than existing MSA-based gLMs), allowing it to capture both across-species evolutionary relationships and within-species longer-range interactions. Trained on whole-genome alignments from 100 vertebrates, RAGenome substantially improves the long-range capabilities of MSA-based gLMs, raising gene finding performance from 0.45 to 0.60, while remaining competitive on purely evolutionary-based tasks like prioritizing pathogenic variants. RAGenome provides competitive gLM performance at a fraction of the training cost, unifying evolutionary modeling and long-range capabilities within a single, flexible, scalable framework. Code is available at https://github.com/PanosAntoniadis/RAGenome.

Thu 8 OctMachine Learning
The gist
Understanding DNA sequences is important for biology and medicine, but analyzing long stretches of DNA is hard for existing language models. The authors present RAGenome, a new model that can read and analyze very long genome sequences by using retrieval methods. This allows it to find genes and understand long-range DNA interactions better than previous models. RAGenome works efficiently and performs well on various tasks involving genomes from many species.
Open → 2610.11761v1

Photoplethysmography model learns features from short to long signals

RAE-PPG: Duration-Grounded Retain-and-Extend Pretraining for PPG Foundation Models

Abstract: Signal features derived from photoplethysmography (PPG) require different signal durations to characterize. Existing PPG foundation models treat duration as a pretraining or evaluation condition rather than using the different durations required by PPG features to organize self-supervision. We hypothesize that self-supervision should expand with signal duration, allowing a single encoder to progressively acquire additional features while preserving and reusing earlier learning. We introduce Retain-and-Extend PPG (RAE-PPG), which trains a single Transformer encoder successively on 10 s, 30 s, and 240 s inputs, adding supervision for signal features supported by each longer observation. The encoder is partitioned into duration-specific parameter groups, allowing later stages to reuse earlier groups while updating only the group assigned to the current stage. Selected earlier targets are reused to supervise later stages, encouraging the corresponding features to remain accessible in longer-input representations. Direct decoding from the final encoder shows that earlier features remain recoverable from longer-input representations, while later-stage features show higher mean decoding performance at their introduction durations. Controlled comparisons further show that prior-stage learning provides a better basis for learning newly introduced features at both transitions. Across 18 tasks from eight datasets, the final frozen encoder achieves the best observed score on 12 tasks compared with five existing PPG foundation models.

Tue 29 SeptMachine Learning
The gist
Different features of pulse signals from photoplethysmography (PPG) need varying amounts of time to measure. The authors propose a method where a single model learns features progressively from short to long signal durations, reusing what it learned before while adding new information. They train a special Transformer model on progressively longer signals (10, 30, and 240 seconds), ensuring that features from shorter signals stay accessible when analyzing longer ones. Their approach improves performance on many tasks compared to other models.
Open → 2609.36794v1

AI agents struggle to reliably analyze data and generate novel scientific hypotheses

DISCERN: Can AI Agents Work Like Scientists and Guide Discovery?

Abstract: Reliable automated research requires agents to vet data, verify analyses, and generate hypotheses grounded in trustworthy evidence, potentially reducing routine scientific workload while allowing scientists to focus on interpretation and discovery. Existing benchmarks often only assess analytical task completion or hypothesis generation separately rather than testing whether reliable evidence supports valid and novel claims. We introduce DISCERN (Data Integrity and Scientific Capability: Evidence, Reasoning, and Novelty), a controlled benchmark on real, publicly available datasets that evaluates three key levels of an automated research workflow. The first two levels test data integrity and analysis verification under confounds and tool traps, while the third tests hypothesis generation and revision under adversarial review, including counterfactual cases in which evidence consistent with real data and documented scientific phenomena conflicts with established expectations, motivating alternative explanations and testable hypotheses. Across 203 tasks, eight life-science tracks, and eight models, DISCERN shows that strong aggregate performance can mask level-specific weaknesses. Agents earn perfect scores in only 60.8% of Level 1, 34.2% of Level 2, and 0.6% of Level 3 evaluations, with penalties attributed to rejection of sound data, failure to carry recognized limitations into conclusions, and wide variation in hypothesis production. Cross-track rankings by token and code use are substantially more stable than rankings by evidence judgment, suggesting greater consistency in computational effort than in evidence-based reasoning. These profiles identify opportunities for supervised scientific assistance, but current agents do not yet demonstrate reliable autonomous analysis or discovery. Code and data: https://huggingface.co/datasets/discern-bench-anon/discern-benchmark

Sun 27 SeptArtificial Intelligence
The gist
Automated AI systems that do scientific research need to check if data is trustworthy, verify their analysis, and create new ideas based on real evidence. The authors created a benchmark called DISCERN to test all these steps together using real public datasets. They found that while AI can do some parts well, it often rejects good data, misses problem details when drawing conclusions, and produces inconsistent new hypotheses. Therefore, current AI agents are not yet able to perform scientific discovery on their own reliably.
Open → 2609.33357v1

Federated learning improves missing clinical data filling across centers

Fed-ReMasker: Federated Tabular Imputation under Feature-Level Missingness

Abstract: Multi-center clinical studies and biomedical research collaborations increasingly seek to utilize data across centers to build models that generalize beyond any single center. This creates two distinct challenges: data protection regulations may restrict the sharing of raw patient data across institutions, while centers may collect only partially overlapping sets of features under different protocols. Federated learning enables collaborative model training without centralizing raw data. However, existing federated imputation methods rarely evaluate feature-level missingness, in which entire features are unobserved at some centers. To address this setting, we adapt the ReMasker masked autoencoder to federated learning (Fed-ReMasker), enabling centers to impute features never observed locally by leveraging knowledge learned across collaborating centers. We evaluate Fed-ReMasker in a benchmark spanning synthetic datasets with linear and nonlinear relationships and real-world tabular datasets, including clinical data. The benchmark varies the number of centers, the missingness ratios, and client heterogeneity. Fed-ReMasker achieves the lowest imputation error in 93.2% of value-level and 96.7% of feature-level scenarios in the homogeneous benchmark. It also remains robust to client heterogeneity using simple federated averaging, outperforming all baselines in all 36 value-level scenarios and each baseline in at least 35 of 36 feature-level scenarios, and comes within 3.0% on average of a centralized model trained on the pooled data.

Wed 23 SeptMachine LearningArtificial Intelligence
The gist
When hospitals and research centers collect different types of patient data but want to use it together, sharing raw data is often not allowed and some data features may be entirely missing in some places. The authors created Fed-ReMasker, a tool that helps predict and fill in missing data without sharing sensitive patient information, by learning patterns from all involved centers. Their tests show that this method guesses missing values more accurately than other existing methods, even when some centers have very different data. It works nearly as well as if all data were combined in one place, which isn’t usually possible.
Open → 2609.28105v1

Biomedical entity linking improves with detailed token matching

BELXTR: Biomedical Entity Linking via Contextualized Token Retrieval

Abstract: Biomedical Entity Linking disambiguates mentions to entities in a knowledge base (KB), making it the cornerstone of information extraction pipelines. While embedding-based models are a popular approach for the task, they suffer from a key limitation. They compress mentions (and entities) into a single vector, forcing the model to average away crucial fine-grained differences. We present BELXTR, a novel embedding model based on the multi-vector (a.k.a. late interaction) architecture, which allows to leverage token-level matching information. BELXTR extends the original XTR model to biomedical entity linking by integrating an existing task-specific training objective and exploring active query expansion. Experiments across ten corpora and five KBs show that BELXTR improves upon current state-of-the-art in half of the corpora with an average improvement of 5pp recall@1. The largest gains are reported on the challenging cross-species gene disambiguation subtask, where BELXTR outperforms an LLM-powered retrieve-and-rerank pipeline and closely approaches a specialized rule-based system. Our results highlight multi-vector models as a practical alternative to hard-to-maintain rule-based systems or in scenarios where LLM-based reranking is too costly as in PubMed-scale mining. The code to reproduce our experiments can be found at: https://github.com/sg-wbi/belxtr.

Tue 22 SeptComputation and Language
The gist
Biomedical entity linking helps computers figure out exactly which medical terms in text refer to which specific entries in a biomedical knowledge database. The authors found that many current methods oversimplify this task by turning each term into a single summary, losing important details. They created a new method called BELXTR that looks at individual word parts within the terms, improving accuracy. Their tests showed it works better on many datasets, especially when distinguishing similar genes across species.
Open → 2609.25859v1

Single-cell models reveal challenges with rare cell type recognition

Rethinking Class Imbalance for Single-Cell Foundation Models: A Systematic Benchmark Across Architectures and Long-Tail Loss Functions

Abstract: Single-cell foundation models (scGPT, scBERT, Geneformer) achieve cell-type classification accuracy up to 97.5% in our experiments, yet this aggregate accuracy can mask systematic failure on rare, often disease-relevant cell populations that long-tail loss functions are widely assumed to address. We present a systematic benchmark of six long-tail loss functions (cross-entropy, weighted CE, class-balanced loss, focal loss, LDAM, logit-adjusted softmax) across three architectures and three datasets (Multiple Sclerosis, Zheng68K, human Pancreas), totaling 162 controlled training runs (3 backbones x 3 datasets x 6 losses x 3 seeds). The gap between overall accuracy, Macro-F1, and rare-class recall under plain cross-entropy is consistent across all nine (architecture, dataset) settings, driven by dataset structure rather than pretraining. Rare-class failure itself splits into two regimes with distinct embedding-geometry signatures, visible before any loss is chosen: some classes are recoverable by the right loss, while others retain linear separability yet are absorbed into unrelated classes' neighborhoods under every evaluated loss and architecture. Among the recoverable classes, the efficacy of reweighting is predicted by a class's absolute training-set size, rather than its share of the dataset or the dataset's overall imbalance ratio. Class-balanced loss and LDAM are the most consistent choices across all nine settings, while logit adjustment trades rare-class precision for recall rather than improving both. Our results give both a reusable benchmark and mechanism-grounded practical guidelines for combining foundation models with imbalanced biological data.

Sun 20 SeptMachine Learning
The gist
Classifying cell types from biological data is easier for common cells but tougher for rare ones, which may be important in disease. The authors tested different methods designed to help detect these rare cells across various models and datasets. They found that some rare cell types can be identified better with certain loss functions, but others remain hard to separate correctly. Also, the usefulness of techniques that give more weight to rare classes depends on the actual number of rare training examples, not just their proportion. This work helps guide better use of AI models with imbalanced biological data.
Open → 2609.23325v1

Scientific knowledge graphs improve distributed vector search efficiency

COMPASS: Steering Distributed Vector Search with Scientific Knowledge Graphs

Abstract: Vector databases use hashing to partition data across "shards," logical units for distributed execution. This placement, however, destroys semantic locality, forcing each query into scatter-gather limited by the slowest shard. Vector-space clustering can help, but scientific evidence is often connected by factual relations that do not align with embedding distance. We present COMPASS, a framework that uses a knowledge graph (KG) to determine data placement and query-time shard selection. COMPASS detects communities, splits oversized communities, inserts embeddings by subject entity, and routes queries to a small set of shards. Across four biomedical KGs, our method searches only 13-18% of the corpus while preserving broadcast recall and recovering up to 2.6x more multi-hop evidence than an embedding-based baseline. On 15 HPC nodes, COMPASS sustains 7.9x higher throughput with lower tail latency than hash-based broadcast. These results show that KG structure provides a compact complement to embedding geometry for scalable vector search.

Fri 11 SeptDatabasesDistributed, Parallel, and Cluster Computing
The gist
Searching large collections of data with vector databases can be slow because data is split randomly across many parts, causing queries to check everything. The authors show that using scientific knowledge graphs, which capture relationships between facts, can better group data and guide queries to fewer relevant parts. Their approach, COMPASS, checks less data while still finding the important connections between pieces of information. This leads to much faster searching without losing important results.
Open → 2609.13452v1