Patterns, a Cell Press journal

A call to join a collective effort on AI evaluation AI evaluations increasingly shape deployment, governance, and trust, but expectations for how they are conducted and reported remain fragmented. We introduce a cross-sector Delphi process to develop community-endorsed guidance for AI evaluation practice and invite researchers, practitioners, ethics organizations, and institutions to participate.

Online Now: A call to join a collective effort on AI evaluation #datascience

Modeling of longitudinal immune profiles reveals distinct immunogenic signatures following five COVID-19 vaccinations among people living with HIV Using machine learning to map immune interdependencies, this study deploys random forests (RFs) to compare vaccine-induced immunity in people with HIV versus HIV-negative age-matched controls across up to five SARS-CoV-2 doses. RFs reveal distinct saliva- and blood-based humoral patterns and a subset of subjects whose responses resembled controls, signaling partial immunologic restoration. Privacy-preserving synthetic patients enable RF training on synthetic data that generalize to real individuals, supporting precision-guided vaccination and follow-up.

Online Now: Modeling of longitudinal immune profiles reveals distinct immunogenic signatures following five COVID-19 vaccinations among people living with HIV #datascience

TweetyBERT: Automated parsing of birdsong through self-supervised machine learning Parsing birdsong into behavioral units typically requires human-labeled data or pre-segmented audio. TweetyBERT, a self-supervised transformer, overcomes these limitations by learning directly from raw spectrograms. Operating at a 2.7 ms temporal resolution, the model preserves a one-to-one correspondence between input time bins and latent states. Applied to canary song, TweetyBERT autonomously discovers syllable-level representations that align closely with expert annotations, enabling large-scale automated labeling of vocal sequences with minimal human intervention.

Online Now: TweetyBERT: Automated parsing of birdsong through self-supervised machine learning #datascience

Promoting sustainable human mobility for income segregation mitigation Despite extensive studies that have addressed the quantification of income segregation, its impact on human mobility remains unclear. This study introduces a segregation visitation index and a mobility prediction model to reveal biased travel patterns, showing how segregation and mobility reinforce inequalities and guide more inclusive, sustainable urban planning.

Online Now: Promoting sustainable human mobility for income segregation mitigation #datascience

DestinyNet: A deep-learning framework for cell-fate analysis from lineage-tracing single-cell RNA sequencing data DestinyNet is a deep-learning framework for analyzing lineage-tracing single-cell sequencing data that integrates fate clustering, fate-flow inference, and fate prediction. It is robust to barcode sparsity and experimental variability, providing a unified approach for interpreting cell-fate dynamics.

Online Now: DestinyNet: A deep-learning framework for cell-fate analysis from lineage-tracing single-cell RNA sequencing data #datascience

Cell Press: A dedication to quality. www.cell.com/about?utm_so...

Creating strong predictive models in oncology Many oncology predictive models fail to improve care. Issues include risks of bias, underpowered radiomics studies, and limited clinical impact. A path forward involves an emphasis on clinically actio...

Also in this issue, check out the Opinion from Michael Gensheimer, who argues that predictive model research in oncology must focus more on generalizable performance and clinical usefulness if patient care is to be improved.
www.cell.com/patterns/ful...

On the cover: Inspired by Jia et al. in this issue, the cover visualises how incomplete or incorrect segmentation labels can be filled in with information from the image itself. The grid marks are where labels are absent, while the completed shapes illustrate how visual context helps recover those regions. The robotic hand represents an automated process that reduces the need for manual relabelling. Overall, the image shows how combining images with partial labels can produce more reliable training data for semantic segmentation. Image credit: Phillip Krzeminski.

Our February issue is now live!
www.cell.com/patterns/iss...

On the cover this month, we are highlighting a work from Jia et al that describes a method to fix defective annotations in image libraries and thereby advance computer vision research.
www.cell.com/patterns/ful...

BDI-Kit: An AI-powered toolkit for biomedical data harmonization BDI-Kit is an extensible open-source toolkit for harmonizing biomedical datasets. By combining automated methods with AI-assisted guidance, it allows researchers and practitioners to interactively identify and refine how dataset fields and their values correspond across datasets, producing consistent and unified datasets ready for downstream analysis.

Online Now: BDI-Kit: An AI-powered toolkit for biomedical data harmonization #datascience

Evolving reservoir computers reveal bidirectional coupling between predictive power and emergent dynamics Biological and artificial neural networks perform complex computations that single neurons cannot. How this collective ability arises remains unclear. Here, the authors study this using bio-inspired reservoir computers and quantify emergence—when a system is more than the sum of its parts. The results reveal a direct, bidirectional link between a network’s predictive power and its emergent dynamics, suggesting emergence as a core mechanism for computation.

Online Now: Evolving reservoir computers reveal bidirectional coupling between predictive power and emergent dynamics #datascience

Experimental and machine learning-based exploration of repurposed drugs reveals chemical features underlying phospholipidosis Phospholipidosis (PLD) is a cellular adverse effect of excessive accumulation of phospholipids that can be triggered by many cationic amphiphilic drugs, therefore bringing challenges for drug discovery. Here, over 5,000 repurposed drugs are tested for PLD induction across multiple cell lines. A robust machine learning model was developed and validated to accurately predict PLD-inducing compounds, offering a powerful early-screening tool to accelerate drug discovery and minimize costly late-stage failures.

Online Now: Experimental and machine learning-based exploration of repurposed drugs reveals chemical features underlying phospholipidosis #datascience

UbiQTree: Uncertainty quantification in XAI with tree ensembles In the context of using AI in healthcare, XAI is a field that helps us understand how decisions are made. One such method is Shapley additive explanations (SHAP). This research studies the stability of SHAP. The ensemble model is used as a tool to quantify uncertainty and study instabilities. The significance of this study is that the most important features are not necessarily the most stable. This research is important for fields such as healthcare, where decisions are critical.

Online Now: UbiQTree: Uncertainty quantification in XAI with tree ensembles #datascience

Variable rate neural compression for sparse detector data High-volume TPC detectors, such as those in the sPHENIX experiment, produce sparse yet information-rich readouts that challenge both storage budgets and discovery potential. The authors present a lightweight sparse-convolution algorithm that picks and saves only the most informative voxels—yielding adaptive compression whose size scales with signal content, not raw tensor size. The method dramatically improves reconstruction quality, raises throughput with increasing sparsity, and thus enables full-dataset retention and real-time compression for next-generation collider experiments.

Online Now: Variable rate neural compression for sparse detector data #datascience

Prewired static visual receptive fields for environment-agnostic perception Deep neural networks often struggle with domain shifts whereas biological brains can flexibly adapt to changing sensory environments. This study shows that static, prewired receptive fields in early visual layers enable robust object recognition through domain-general, shape-based representations.

Online Now: Prewired static visual receptive fields for environment-agnostic perception #datascience

Hierarchical fusion of electrocardiogram and phonocardiogram data improves heart failure detection By fusing electrical and acoustic heart signals, a local-to-global hierarchical fusion network detects heart failure with reduced ejection fraction at the point of care. With a new multi-channel digital stethoscope dataset and superior accuracy, this study highlights an accurate, non-invasive, scalable tool for early heart failure screening.

Online Now: Hierarchical fusion of electrocardiogram and phonocardiogram data improves heart failure detection #datascience

Embeddings from language models are good learners for single-cell data analysis scELMo introduces a way to analyze massive single-cell datasets by harnessing large language models to summarize biological knowledge about each gene. These summaries are transformed into mathematical embeddings that integrate with cell data, enabling efficient cell grouping, batch correction, and treatment prediction. By merging language understanding with biological data, scELMo reduces computational demands and democratizes advanced analysis—offering a faster, more accessible path to discoveries that could inform new therapies and deepen understanding of human health.

Online Now: Embeddings from language models are good learners for single-cell data analysis #datascience

Check out also the related news coverage via PlumX plu.mx/plum/a/?doi=...
or Altmetric www.altmetric.com/details/1858...

An image of the first page of the paper mentioned in the post.

📈In case you missed it, our most read paper for the last month has been:

The carbon and water footprints of data centers and what this could mean for artificial intelligence
doi.org/10.1016/j.pa...

by Alex de Vries of @digiconomist.bsky.social and @vuamsterdam.bsky.social

Recalibrating academic expertise in the age of generative AI GenAI tools promise accelerated discovery but risk eroding scientific expertise. When researchers delegate to AI, they bypass the effortful processes through which competence develops. This perspectiv...

📣 New Perspective 📣

In our article "Recalibrating academic expertise
in the age of generative AI", Zhicheng Lin and I discuss how an over-reliance on generative AI in academia may erode key skills of scientific enquiry.

Out now in @cp-patterns.bsky.social, read it below! 👇

tinyurl.com/py38vnvb

On the cover: The central brain in this image represents IoT-LLM, a framework for processing real-world sensor data with large language models, which is described by An et al. in this issue of Patterns. IoT-LLM provides a central reasoning core that connects and interprets heterogeneous sensor inputs grounded in the physical world such as electrocardiography, temperature, motion, camera images, and robotic system sensors. In the image, these data streams converge on the central model, highlighting how IoT-LLM performs retrieval-augmented fusion and structured reasoning across different data types. Systems like IoT-LLM are laying a foundation for future embodied robots operating in real-world environments. Image credit: An Tuo, Nanyang Technological University.

Our first issue of 2026 is now live!
www.cell.com/patterns/iss...

This month's cover image highlights a framework, IoT-LLM, for applying large language model driven reasoning to real-world sensor data www.cell.com/patterns/ful...

Text on image reads: Patterns, A Cell Press journal, Editors' Pick, Best of 2025, Explore the collection

Check out some of our best papers from 2025, as selected by the journal's editors info.cell.com/collection-r...

IoT-LLM: A framework for enhancing large language model reasoning from real-world sensor data IoT-LLM enables language models to understand and reason about the physical world through real-world sensors. By converting diverse signals, such as motion and electrocardiography data, into meaningful insights, it bridges data and understanding, advancing explainable and human-aware intelligence for future embodied AI.

Online Now: IoT-LLM: A framework for enhancing large language model reasoning from real-world sensor data #datascience

A label masked autoencoder for image-guided segmentation label completion Many segmentation datasets carry gaps or noise in their annotations, which blunts model training. Here, the authors present a label-image fusion approach that learns to fill missing or corrupted regions. By turning imperfect labels into dependable supervision, it upgrades existing datasets and lifts accuracy without fresh hand labeling. The idea offers a simple, scalable approach to maintaining and expanding datasets across benchmarks and application domains.

Online Now: A label masked autoencoder for image-guided segmentation label completion #datascience

Covered by @gizmodo.com gizmodo.com/ai-image-gen...

Online Now: Autonomous language-image generation loops converge to generic visual motifs #datascience

AI’s water and electricity use soars in 2025 It’s guzzling up even more water than expected.

Check out the coverage @theverge.com www.theverge.com/news/845831/...

The carbon and water footprints of data centers and what this could mean for artificial intelligence Company-wide metrics from the environmental disclosure of data center operators suggest that AI systems may have a carbon footprint equivalent to that of New York City in 2025, while their water footprint could be in the range of the global annual consumption of bottled water. Further disclosures from data center operators are urgently required to improve the accuracy of these estimates and to responsibly manage the growing environmental impact of AI systems.

Online Now: The carbon and water footprints of data centers and what this could mean for artificial intelligence #datascience

Online Now: Mainzelliste: Ten years of pseudonymization, record linkage, and informed consent management #datascience

On the cover: The image, related to the work by Fernandez Bonet et al., shows a series of spatially coherent networks, explored with breadth-first search to differing depths. The emerging predictable scaling pattern revealed by the progression is indicative of geometric consistency, and this feature can be detected in networks as an indication of how “spatial” a network is. Image credit to Ian Hoffecker and David Fernandez Bonet.

Our December issue is live!
www.cell.com/patterns/iss...

On the cover this month is an image highlighting the work by David Fernandez Bonet and co-authors, which introduces a set of geometry-based metrics for assessing the quality of DNA barcode networks
www.cell.com/patterns/ful...

Node persistence from topological data analysis reveals changes in brain functional connectivity This study employs persistent homology to investigate alterations in brain functional connectivity associated with healthy aging and autism spectrum disorder (ASD). Node persistence, a scalable local measure based on persistent homology introduced in this study, identifies brain regions linked to these conditions, including those with clinical evidence from non-invasive brain stimulation.

Online Now: Node persistence from topological data analysis reveals changes in brain functional connectivity #datascience