Biology subjects

Koo, P.

Publications and source records attributed to Koo, P..

13 recordsLinked to original sources

bioRxiv2026

Dissecting and directing pathology foundation models

Foundation models (FMs) are central to digital pathology, encoding histology images into dense embeddings for facilitating diagnostic classification, molecular alteration prediction, and clinical outcome modeling. However, the opacity of these embeddings renders FM-based systems "black boxes," limiting their trustworthiness for clinical translation and utility for scientific discovery. Here, we introduce PICASSO (Pathology Image Concept Atlas built via SparSe dictiOnary learning), a framework that makes pathology FMs interpretable and controllable. PICASSO decomposes FM embeddings into human-interpretable visual concepts using a sparse autoencoder. It is trained on more than 120 million tissue patches across 32 cancer types, producing the first pan-cancer atlas of histomorphological concepts. We demonstrate that PICASSO enables diverse downstream applications of FM embeddings by exposing interpretable structure within learned representations and supporting concept-level intervention. It enables auditing of clinical model behavior by revealing the morphological features driving predictions. Beyond transparency and validation, PICASSO enables the discovery of new biological insights; for example, it identified hobnailing epithelial morphology as a previously unrecognized biomarker of EGFR mutations in lung adenocarcinoma. By linking PICASSO-derived concepts with spatial transcriptomics, we uncover associations between morphological patterns and gene expression programs. Furthermore, PICASSO allows suppression of concepts associated with technical artifacts, thereby reducing model reliance on spurious signals. Finally, PICASSO enables controlled manipulation of learned concepts to generate counterfactual embeddings for exploratory therapeutic analysis, such as modulating tumour-infiltrating lymphocyte density to assess impacts on predict survival outcomes. Together, PICASSO provides a principled framework for transforming pathology FMs into platforms for mechanistic insight and discovery.

pathology↗

bioRxiv2025

A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction

Genomic prediction and design require models that integrate local sequence features with long-range regulatory dependencies spanning hundreds of kilobases to megabases. Existing approaches have made substantial progress along complementary axes: supervised sequence-to-function models achieve high accuracy for specific assays and organisms, self-supervised genomic foundation models learn transferable representations from large-scale sequence data, and conditional generative models enable principled sequence design guided by functional objectives. However, these strengths are typically realized in isolation--across distinct model classes, architectures, and training regimes--limiting the ability to combine long-context, base-resolution prediction, functional modeling, and controllable generation within a single efficient framework that generalizes across organisms and modalities. Here we introduce Nucleotide Transformer v3 (NTv3), a multi-species foundation model that unifies representation learning, functional-track and genome-annotation prediction, and controllable sequence generation within a common backbone. NTv3 uses a U-Net-like architecture to enable single-base tokenization and efficient modeling of contexts up to 1 Mb. We pre-train NTv3 on 9 trillion base pairs from OpenGenome2 using base-resolution masked language modeling, followed by post-training with a joint objective that integrates continued self-supervision with supervised learning on [~]16,000 functional tracks and annotation labels from 24 animal and plant species. After post-training, NTv3 achieves state-of-the-art accuracy for functional-track prediction and genome annotation across species, outperforming leading sequence-to-function and foundation-model baselines on established benchmarks and on the new NO_SCPLOWTVC_SCPLOW 3 BO_SCPLOWENCHMARKC_SCPLOW, a controlled downstream fine-tuning suite in a standardized 32 kb input / base-resolution output setting. We further show that NTv3 consolidates a shared regulatory grammar across tasks, enabling coherent long-range genome-to-function inference and variant-associated remodeling. Finally, we fine-tune NTv3 into a controllable generative model via masked diffusion language modeling and use it to design enhancer sequences with specified activity levels and promoter selectivity. We validate these designs experimentally, showing that generated enhancers recapitulate the intended activity stratification and achieve the desired promoter-specific activation in cellulo. We release the NTv3 model family together with code and practical cookbooks for long-context training, multispecies post-training, fine-tuning, interpretation, and sequence design.

genomics↗

bioRxiv2025

Uncovering the Mechanistic Landscape of Regulatory DNA with Deep Learning

The regulatory genome encodes the logic that governs gene expression, enabling cells to respond to developmental, environmental, and evolutionary cues. This logic arises from complex cis-regulatory mechanisms that integrate transcription factor motifs, their syntactical arrangement, and surrounding sequence context, features that remain challenging to decode. Here, we present SEAM (Systematic Explanation of Attribution-based Mechanisms), a computational framework that combines deep learning with explainable AI to map the mechanistic impact of genetic mutations. Applied to human and Drosophila regulatory loci, SEAM uncovers functional binding sites at sequences of interest and identifies which mutations preserve, disrupt, or create novel binding sites. SEAM also reveals that two qualitatively distinct classes of regulatory signal are operative at many loci: signals that are robust to mutation and signals that are readily reprogrammable. These results clarify the inherent ability of regulatory DNA to evolve. They also position SEAM as a versatile framework for interpreting non-coding variants and for informing the mechanism-aware design of synthetic sequences.

genomics↗

bioRxiv2025

GAME: Genomic API for Model Evaluation

The rapid expansion of genomics datasets and the application of machine learning has produced sequence-to-activity genomics models with ever-expanding capabilities. However, benchmarking these models on practical applications has been challenging because individual projects evaluate their models in ad hoc ways, and there is substantial heterogeneity of both model architectures and benchmarking tasks. To address this challenge, we have created GAME, a system for large-scale, community-led standardized model benchmarking on user-defined evaluation tasks. We borrow concepts from the Application Programming Interface (API) paradigm to allow for seamless communication between pre-trained models and benchmarking tasks, ensuring consistent evaluation protocols. Because all models and benchmarks are inherently compatible in this framework, the continual addition of new models and new benchmarks is easy. We also developed a Matcher module powered by a large language model (LLM) to automate ambiguous task alignment between benchmarks and models. Containerization of these modules enhances reproducibility and facilitates the deployment of models and benchmarks across computing platforms. By focusing on predicting underlying biochemical phenomena (e.g. gene expression, open chromatin, DNA binding), we ensure that tasks remain technology-independent. We provide examples of benchmarks and models implementing this framework, and anticipate that the community will contribute their own, leading to an ever-expanding and evolving set of models and evaluation tasks. This resource will accelerate genomics research by illuminating the best models for a given task, motivating novel functional genomic benchmarks, and providing a more nuanced understanding of model abilities.

bioinformatics↗

bioRxiv2024

Interpretably deep learning amyloid nucleation by massive experimental quantification of random sequences

Protein aggregation is a pathological hallmark of more than fifty human diseases and a major problem for biotechnology. Methods have been proposed to predict aggregation from sequence, but these have been trained and evaluated on small and biased experimental datasets. Here we directly address this data shortage by experimentally quantifying the amyloid nucleation of >100,000 protein sequences. This unprecedented dataset reveals the limited performance of existing computational methods and allows us to train CANYA, a convolution-attention hybrid neural network that accurately predicts amyloid nucleation from sequence. We adapt genomic neural network interpretability analyses to reveal CANYAs decision-making process and learned grammar. Our results illustrate the power of massive experimental analysis of random sequence-spaces and provide an interpretable and robust neural network model to predict amyloid nucleation.

genomics↗

bioRxiv2024

Designing DNA With Tunable Regulatory Activity Using Discrete Diffusion

Designing regulatory DNA with tunable and context-specific activity is a major goal in biotechnology and medicine. Deep generative models offer a promising route for sequence design, yet it remains unclear whether synthetic sequences faithfully recapitulate the motif organization and functional specificity of natural regulatory DNA. Here we present DNA Discrete Diffusion (D3), a generative model that designs regulatory DNA through an iterative nucleotide-substitution process. Across computational benchmarks, D3 improves regulatory sequence generation relative to matched diffusion baselines, producing sequences that more closely match target activity, activity distributions, and sequence composition. In K562 lentiMPRA experiments, D3-designed sequences retained measurable regulatory activity and more closely recapitulated the activity distribution of genomic regulatory sequences than matched diffusion baselines. D3 performs robustly with limited training data and generates sequences informative enough to improve predictive models when labeled data are scarce. When trained without activity labels on task-specific regulatory sequence sets, D3 learns frozen sequence representations that are predictive of enhancer activity and compare favorably to several off-the-shelf genomic language model embeddings. Analysis of the sampling process identifies reproducible phases of sequence exploration, compositional refinement, and motif-associated convergence, providing an interpretable view of how D3 constructs regulatory sequences. These results establish D3 as a practical framework for designing synthetic regulatory DNA and studying sequence features associated with context-specific regulatory activity.

genomics↗

bioRxiv2024

EvoAug-TF: Extending evolution-inspired data augmentations for genomic deep learning to TensorFlow

Deep neural networks (DNNs) have been widely applied to predict the molecular functions of regulatory regions in the non-coding genome. DNNs are data hungry and thus require many training examples to fit data well. However, functional genomics experiments typically generate limited amounts of data, constrained by the activity levels of the molecular function under study inside the cell. Recently, EvoAug was introduced to train a genomic DNN with evolution-inspired augmentations. EvoAug-trained DNNs have demonstrated improved generalization and interpretability with attribution analysis. However, EvoAug only supports PyTorch-based models, which limits its applications to a broad class of genomic DNNs based in TensorFlow. Here, we extend EvoAugs functionality to TensorFlow in a new package we call EvoAug-TF. Through a systematic benchmark, we find that EvoAug-TF yields comparable performance with the original EvoAug package. AvailabilityEvoAug-TF is freely available for users and is distributed under an open-source MIT license. Researchers can access the open-source code on GitHub (https://github.com/p-koo/evoaug-tf). The pre-compiled package is provided via PyPI (https://pypi.org/project/evoaug-tf) with in-depth documentation on ReadTheDocs (https://evoaug-tf.readthedocs.io). The scripts for reproducing the results are available at (https://github.com/p-koo/evoaug-tf_analysis).

bioinformatics↗

bioRxiv2023

Light and temperature regulate m6A-RNA modification to regulate growth in plants

N6-methyladenosine is a highly dynamic, abundant mRNA modification which is an excellent potential mechanism for fine tuning gene expression. Plants adapt to their surrounding light and temperature environment using complex gene regulatory networks. The role of m6A in controlling gene expression in response to variable environmental conditions has so far been unexplored. Here, we map the transcriptome-wide m6A landscape under various light and temperature environments. Identified m6A-modifications show a highly specific spatial distribution along transcripts with enrichment occurring in 5UTR regions and around transcriptional end sites. We show that the position of m6A modifications on transcripts might influence cellular transcript localization and the presence of m6A-modifications is associated with alternative polyadenylation, a process which results in multiple RNA isoforms with varying 3UTR lengths. RNA with m6A-modifications exhibit a higher preference for shorter 3UTRs. These shorter 3UTR regions might directly influence transcript abundance and localization by including or excluding cis-regulatory elements. We propose that environmental stimuli might change the m6A landscape of plants as one possible way of fine tuning gene regulation through alternative polyadenylation and transcript localization.

plant biology↗

bioRxiv2022

Evaluating deep learning for predicting epigenomic profiles

Deep learning has been successful at predicting epigenomic profiles from DNA sequences. Most approaches frame this task as a binary classification relying on peak callers to define functional activity. Recently, quantitative models have emerged to directly predict the experimental coverage values as a regression. As new models continue to emerge with different architectures and training configurations, a major bottleneck is forming due to the lack of ability to fairly assess the novelty of proposed models and their utility for downstream biological discovery. Here we introduce a unified evaluation framework and use it to compare various binary and quantitative models trained to predict chromatin accessibility data. We highlight various modeling choices that affect generalization performance, including a downstream application of predicting variant effects. In addition, we introduce a robustness metric that can be used to enhance model selection and improve variant effect predictions. Our empirical study largely supports that quantitative modeling of epigenomic profiles leads to better generalizability and interpretability.

bioinformatics↗

bioRxiv2022

Statistical correction of input gradients for black box models trained with categorical input features

Post-hoc attribution methods are widely applied to provide insights into patterns learned by deep neural networks (DNNs). Despite their success in regulatory genomics, DNNs can learn arbitrary functions outside the probabilistic simplex that defines one-hot encoded DNA. This introduces a random gradient component that manifests as noise in attribution scores. Here we demonstrate the pervasiveness of off-simplex gradient noise for genomic DNNs and introduce a statistical correction that is effective at improving the interpretability of attribution methods.

bioinformatics↗

bioRxiv2020

Single Layers of Attention Suffice to Predict Protein Contacts

AO_SCPLOWBSTRACTC_SCPLOWThe established approach to unsupervised protein contact prediction estimates co-evolving positions using undirected graphical models. This approach trains a Potts model on a Multiple Sequence Alignment, then predicts that the edges with highest weight correspond to contacts in the 3D structure. On the other hand, increasingly large Transformers are being pretrained on protein sequence databases but have demonstrated mixed results for downstream tasks, including contact prediction. This has sparked discussion about the role of scale and attention-based models in unsupervised protein representation learning. We argue that attention is a principled model of protein interactions, grounded in real properties of protein family data. We introduce a simplified attention layer, factored attention, and show that it achieves comparable performance to Potts models, while sharing parameters both within and across families. Further, we extract contacts from the attention maps of a pretrained Transformer and show they perform competitively with the other two approaches. This provides evidence that large-scale pretraining can learn meaningful protein features when presented with unlabeled and unaligned data. We contrast factored attention with the Transformer to indicate that the Transformer leverages hierarchical signal in protein family databases not captured by our single-layer models. This raises the exciting possibility for the development of powerful structured models of protein family databases.1

bioinformatics↗

bioRxiv2020

The structure-fitness landscape of pairwise relations in generative sequence models

If disentangled properly, patterns distilled from evolutionarily related sequences of a given protein family can inform their traits - such as their structure and function. Recent years have seen an increase in the complexity of generative models towards capturing these patterns; from sitewise to pairwise to deep and variational. In this study we evaluate the degree of structure and fitness patterns learned by a suite of progressively complex models. We introduce pairwise saliency, a novel method for evaluating the degree of captured structural information. We also quantify the fitness information learned by these models by using them to predict the fitness of mutant sequences and then correlate these predictions against their measured fitness values. We observe that models that inform structure do not necessarily inform fitness and vice versa, contrasting recent claims in this field. Our work highlights a dearth of consistency across fitness assays as well as divergently provides a general approach for understanding the pairwise decomposable relations learned by a given generative sequence model.

bioinformatics↗

bioRxiv2019

N-WASP regulates the mobility of the B cell receptor and co-receptors during signaling activation

Regulation of membrane receptor mobility is important in tuning the cells response to external signals. This is particularly relevant in the context of immune receptor signaling. The binding of B cell receptors (BCR) to antigen induces B cell receptor activation. While actin dynamics and BCR signaling are known to be linked, the role of actin dynamics in modulating receptor mobility is not well understood. Here, we use single molecule imaging to examine BCR movement during signaling activation and examine the role of actin dynamics on BCR mobility. We use a novel machine learning based method to classify BCR trajectories into distinct diffusive states and show that the actin regulatory protein N-WASP regulates receptor mobility. Constitutive loss or acute inhibition of N-WASP, which is associated with enhanced signaling, leads to a predominance of BCR trajectories with lower diffusivity and is correlated with a decrease in actin dynamics. Furthermore, loss of N-WASP reduces diffusivity of CD19, a stimulatory co-receptor of the BCR but not that of unstimulated Fc{gamma}RIIB, an inhibitory co-receptor. The effect of N-WASP is mirrored by inhibition of the Arp2/3 complex and formins. Our results implicate the dynamic actin network in fine-tuning receptor mobility and receptor-ligand interactions, thereby modulating B cell signaling.

biophysics↗