(Created with Midjourney 8.1)
□ Proto: A high-level programming language for generative biology
https://www.biorxiv.org/content/10.64898/2026.06.22.733870v1
Proto encodes generative design campaigns across diverse modalities and scales—spanning DNA, RNA, proteins, ligands, and their interactions. Proto readily incorporates predictive models into generative workflows, leveraging them to design alternatively spliced introns with experimental validation in human cell lines.
Proto defines modularity and compositionality at a functional or semantic level and uses generative modeling to bridge this abstractiontolow-levelbiologicalsequences, while preserving global functional coherence.
□ Omnii: A new frontier in generative genomics with Omnii
https://www.radicalnumerics.ai/blog/omnii-health-preview
Omnii is the first genomic Language Model to be mid/post-trained, unlocking new applications without requiring training of specialized classifiers on frozen embeddings. Omnii can reason over DNA tokens, genomic annotation tokens, and other general-purpose special tokens.
Omnii models are pretrained with native fusion of information-dense genomic annotation tracks alongside DNA sequence and are post-trained to be directly applicable to downstream design and prediction tasks. Omnii ingests DNA sequence and conservation signals.
Omnii features a new architecture with a multi-hybrid backbone featuring block convolutions (a generalization of scalar convolutions, suited to genomic modeling), a dynamic sparse attention mechanism, and a 2 million context window.
□ CellOS: Learning a World Model of Cellular State through Joint Embedding Prediction
https://www.biorxiv.org/content/10.64898/2026.06.18.733163v1
CellOS, a generative multi-view single-cell foundation model that learns cellular representations from paired expression and perception views. CellOS keeps the cell-sentence interface but pairs the conventional expression view with a perception view that re-ranks genes by population-relative surprisal.
CellOS employs a token-balanced data engine to partition distributed training by estimated token load, enabling pretraining of a 12-billion-parameter model on 390.5 million single-cell transcriptomes.
CellOS integrates complementary views through a scalable three-stage training strategy that combines causal cell-sentence language modelling, function-preserving dense-to-mixture-of-experts expansion and latent-space alignment via an LLM-JEPA objective.
□ OmicOS: A Comprehensive Omics Ecosystem Infrastructure and Agent System for the AI Era
https://www.biorxiv.org/content/10.64898/2026.06.11.731775v1
OmicOS is an agent-facing operating layer that exposes a reconstructed omics ecosystem as a state-aware, executable action space rather than asking agents to assemble one-off scripts from scattered documentation.
Omic Verse V2 provides the community substrate: scalable AnnDataOOM-compatible rust backends, agent-friendly Python algorithms for single-cell, spatial, bulk and multi-omics analysis, interfaces to single-cell foundation models.
RebuildR automatically reconstructs and evolves R/Bioconductor methods as Python-native implementations under output-equivalence gates. Once a port is equivalent, RebuildR optimizes it through a self-evolving acceleration loop that cannot break equivalence.
□ BioSymphony: Al Agent Skills For Biological Research
https://biosymphony.github.io/
BioSymphony builds public toolkits for using coding agents in biological research. BioSymphony covers biosynthetic route exploration, structural biology campaigns, small-molecule design, natural-product genome mining, fermentation experiment design, and cryo-EM review.
BioSymphony GeneCluster helps Claude Code, Codex, Symphony, and other agent harnesses work through long comparative-genomics tasks without losing track of inputs, controls, routes, tool runs, and review packets.
BioProspector helps agents and agentic harnesses tackle long-horizon bioprospecting work. The skill gives the agent a campaign contract, ledgers, validators, issue templates, and review artifacts it can use to plan the work and keep later workers aligned.
□ Reasoning as Attractor Dynamics: Latent Memory Retrieval via Gibbs-Weighted Energy Minimization
https://arxiv.org/abs/2606.24543
A retrieval mechanism that employs a Gibbs measure of the trajectory’s spectral entropy. It samples multiple reasoning paths and weights them by their inverse energy (P ∝ e−βE) to approximate the equilibrium distribution of the associative memory.
This model represents the reasoning process not as a series of independent token predictions, but as the evolution of a state vector in a high-dimensional energy landscape.
An inverse-square weighting acts as a critical filter. It induces a phase transition where the probability mass of the “Sharp Minima” evaporates, while the “Flat Minima” retain their mass.
□ scDataset: Scalable Data Loading for Deep Learning on Large-Scale Single-Cell Omics
https://arxiv.org/abs/2506.01883
scDataset, a PyTorch data loader that enables efficient training from on-disk data with seamless integration across diverse storage formats. It combines block sampling and batched fetching to achieve quasi-random sampling that balances I/O efficiency with mini-batch diversity.
On Tahoe-100M, a dataset of 100 million cells, scDataset achieves more than two orders of magnitude speedup compared to true random sampling while working directly with AnnData files.
□ CycleVI: Isolating cell cycle variation with an interpretable deep generative model
https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btag372/8714055
CycleVI, an interpretable deep generative model that disentangles cell cycle-driven transcriptional variation from other sources of variation in static scRNA-seq data.
CycleVI uses a VAE architecture with a partitioned latent space comprising a circular subspace encoding the continuous cell cycle phase and a residual subspace capturing remaining variation. Gene expression is reconstructed using a dual-decoder design w/ a strong inductive bias.
A gene-specific Fourier series decoder models periodic, cell cycle-dependent expression, while a neural network decoder models non-cycling variation. An adversarial classifier enforces separation by discouraging the residual latent space from encoding cell cycle information.
□ FateLimit quantifies the prediction horizon of cell fate
https://www.biorxiv.org/content/10.64898/2026.06.22.733672v1
FateLimit, an information-theoretic framework for measuring the temporal dynamics of cell-fate predictability from single-cell omics data.
FateLimit combines probabilistic fate assignment, fate entropy and mutual information to quantify how information about future cellular outcomes is encoded in present molecular states.
Multipotent progenitors exhibit short prediction horizons and high fate entropy, whereas committed cellular states maintain long prediction horizons and low uncertainty.
□ Stability-driven multi-omics integration for reproducible latent structure
https://www.biorxiv.org/content/10.64898/2026.06.23.734064v1
Using sparse generalized canonical correlation analysis (SGCCA) within a repeated cross-validation framework, the reproducibility of latent component sample scores across resampling is systematically evaluated.
The stability of feature selection, the corresponding sparse weight vectors, and the consistency of associations are assessed using strictly out-of-sample (OOS) projections.
Only a subset of latent components exhibit reproducible biological structures and generalizable disease-associated patterns, whereas higher-order components are often unstable and prone to overinterpretation.
□ CSCN: Inference of Cell-Specific Causal Networks Using Single-Cell RNA-Seq Data
https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btag480/8722299
Cell-Specific Causal Network (CSCN) framework infers directed, cell-specific gene regulatory relationships by explicitly modeling causality. CSCN combines causal discovery techniques with efficient computation using kd-trees and bitmap indexing。
CSCN performs conditional independence testing, yielding sparse and interpretable causal graphs for each cell that effectively suppress indirect and spurious associations.
Across nine scRNA-seq datasets, the Causal Katz Matrix (CKM) derived from CSCN provided more accurate and stable cell-state discrimination than expression-based and network-based baselines.
□ OmicsTransformer: Self-Supervised Masked Consistency and Uncertainty-Aware Fusion for Robust Multi-Omics Prediction
https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btag468/8721561
OmicsTransformer, an end-to-end framework that projects omics modalities into latent patches, enforces masked semantic consistency via an Exponential Cosine Consistency Loss, models patch dependencies w/ a Transformer encoder, and fuses modalities by sample-specific uncertainty.
This objective acts as a structural inductive bias: by compelling the model to reconstruct masked semantics from correlated features, it suppresses independent noise and organizes each modality into a smooth latent manifold that preserves continuous biological gradients.
□ MintCNA: A Unified Framework for Integrative Copy Number Profiling with Single-Cell Multi-Omics Data
https://www.biorxiv.org/content/10.64898/2026.06.26.734559v1
MintCNA, a unified framework for CNA profiling that operates on paired scDNA-seq and scRNA-seq from the same cell or scDNA-seq data alone. MintCNA presents novelty in two aspects.
MinCNA denoises signals through an attention-guided convolutional autoencoder (AG-DCAE) with edge-preserving loss that avoids over-smoothing across breakpoint transitions while preserving the piecewise-constant CNA structure.
MinCNA applies a multivariate change-point procedure to segment the genome by jointly integrating evidence across modalities while accounting for the structured missingness of scRNA-seq data.
□ Glitch genes: embedding geometry predicts functional fragility in single-cell foundation models
https://www.biorxiv.org/content/10.64898/2026.06.22.733850v1
The "SolidGoldMagikarp" phenomenon, in which tokens near the global embedding centroid produce bizarre and non-deterministic completions, demonstrated that representational anomalies are not merely curiosities but indicators of functional fragility.
Gene token embedding matrices are extracted from 3 architecturally distinct foundation models: Geneformer, scGPT, and scFoundation, their geometry characterized, statistical outliers identified, and associations between geometric anomaly and perturbation sensitivity evaluated.
□ ARISE: RNA-Anchored Shared-Edge Topology and Hierarchical Fusion for Spatial Multi-Omics Integration
https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btag465/8721301
ARISE (Anchored RNA for Integrated Spatial Embedding) defines a shared-edge topology by intersecting RNA feature-similarity and spatial-proximity graphs, encodes auxiliary modalities on this common scaffold, and integrates them through inside-out hierarchical fusion.
ARISE integrates dual RNA embeddings and progressively incorporates auxiliary modality embeddings into a unified latent representation. Modality-specific and fused reconstruction objectives, together with spatial coherence regularization, constrain the resulting latent space.
□ SABER: Seed variation impacts clustering stability in Single-Cell RNA-Seq and can be mitigated by StAbility-BasEd-Reassignment
https://www.biorxiv.org/content/10.64898/2026.06.16.732539v1
StAbility-BasEd Reassignment (SABER), a Scanpy-based framework that identifies seed-sensitive cells across repeated clusterings and reassigns them to stable cluster cores using cosine similarity.
SABER effectively balances cluster stability and biological relevance, retaining all cells while improving clustering quality over random assignment and preserving annotation concordance with 3.5-fold lower median memory usage than standard Seurat-Louvain clustering.
□ UniversalEPI: robust prediction of cell type-specific and differential chromatin interactions from DNA sequence and chromatin accessibility
https://academic.oup.com/nar/article/54/10/gkag485/8688745
UniversalEPI, an attention-based deep ensemble model that predicts EPIs up to 2 Mb apart using only DNA sequence and chromatin accessibility (ATAC-seq) data. UniversalEPI focuses on biologically relevant, sparse chromatin interactions between accessible regulatory elements.
The architecture of UniversalEPI only accounts for the information coming from accessible chromatin regions (ATAC-seq peaks); all information about genomic regions between accessible regions is provided to the model via the value of the distance between ATAC-seq peaks.
UniversalEPI employs a linear head to embed auxiliary information. UniversalEPI sets the embedding dimension equal to the transformer’s hidden dimension. The embedded vector is then concatenated with the input embeddings, resulting in a new input to the attention layers.
□ Probing the limits of genetic recoding using multi-omics-guided evolution
https://www.nature.com/articles/s41467-026-74300-9
A multi-omics-guided technology that identifies fitness-decreasing errors in synthetic genomes and enables their debugging via genome editing and directed evolution. They demonstrate that—by integrating multi-omics analyses along all steps of the central dogma.
Synonymous codon changes impact mRNA and protein production, influence translation efficiency, induce promoters, and reduce fitness. Beyond discoveries about genomes and the genetic code, this work provides a technology for resolving defects in synthetic chromosomal regions.
□ NanoCellAnnotator: Formalizing Expert Cell Type Annotation with Large Language Models
https://www.biorxiv.org/content/10.64898/2026.06.21.728965v1
NanoCellAnnotator projects cluster-specific marker genes into functional programs using Gene Ontology (GO) enrichment and GO-slim projection, providing structured biological context without premature cell-type assignment.
NanoCellAnnotator employs a deterministic confidence scoring mechanism that evaluates marker-gene support and lineage separation independently of the language model to explicitly flag ambiguous clusters.
Spatially coherent cellular domains are identified using Hybrid Spatially Regularized Non-negative Matrix Factorization (hSNMF). This stage is model-agnostic and operates independently of any cell-type labels to define geometrically contiguous domains.
□ SimMapNet: a Bayesian framework for gene regulatory network inference using gene ontology similarities as external hint
https://link.springer.com/article/10.1186/s12859-026-06542-9
SimMapNet, a Bayesian framework that estimates the precision matrix, which serves as the adjacency matrix in a Gaussian graphical model for GRN inference.
SimMapNet uses the precision matrix to represent the network structure, where non-zero entries indicate regulatory relationships between genes; the precision matrix is the inverse of the covariance matrix.
SimMapNet adopts a Bayesian approach, incorporating GO similarities into the estimation of the hyperparameters of the prior distribution. GO similarities help to define the prior covariance structure, which guides the Bayesian inference process.
□ scBench-Long: Verifiable Benchmarking of Long-Horizon Single-Cell Biology
https://latch.bio/scbench-long
scBench-Long, a benchmark for long-horizon single-cell biology in which agents must recover scientific conclusions from raw or near-raw data without prescribed methods.
The benchmark stages the data and context needed to test a specific biological claim, then grades the final structured answer deterministically.
This follows prior verifiable long-horizon benchmark design while adapting it to single-cell tasks that require donor-aware reasoning, immune-repertoire analysis, chromatin integration, cross-species mapping, and orthogonal validation evidence.
□ EventHorizon: A Foundation Model for Clinical Flow Cytometry
https://www.biorxiv.org/content/10.64898/2026.06.18.733197v1
EventHorizon, a self-supervised foundation model for clinical flow cytometry that produces unified specimen-level representations from heterogeneous multi-panel data.
EventHorizon employs a two-stage hierarchical transformer architecture with marker-aware tokenization, enabling seamless integration of cells measured across different antibody panels into a single shared latent space.
□ SECTOR: structural entropy-based learning of spatiotemporal organisation in spatial transcriptomics
https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btag367/8708325
SECTOR (Structural Entropy-based Clustering and pseudoTime ORdering) optimises a differentiable structural entropy objective on a fused spatial–expression graph, with spatial total variation regularisation to promote tissue continuity.
SECTOR reformulates it as a differentiable loss on a fused spatial-expression graph, modelled through a graph-encoding tree representing root-clusters-spatial locations.
By minimising structural entropy with spatial total variation regularisation, SECTOR favours low-entropy, spatially coherent domains while retaining the graph connectivity required for pseudotime reconstruction.
□ CAGNet: a structure-aware clustering-alternated graph network for cell-cell interaction inference in spatial transcriptomics
https://link.springer.com/article/10.1186/s12859-026-06534-9
CAGNet (Clustering-Alternated Graph Network) integrates clustering into the core representation learning process through an alternating optimization mechanism. CAGNet iteratively updates node embeddings and cluster centers by fixing one component while optimizing the other.
CAGNet employs a graph attention network (GAT) to encode spatial transcriptomics data into structure-aware embeddings, followed by a soft clustering module to refine cluster assignments.
The clustering structure serves as an intermediate guidance signal that improves the quality of node embeddings, which are subsequently used for CCI link prediction. Meanwhile, a reconstruction module is employed to update embeddings based on graph structure constraints.
□ RepGene: Toward a Unified Gene Representation Space Robust to Missing Biological Views
https://www.biorxiv.org/content/10.64898/2026.06.11.731512v1
RepGene, a lightweight single-branch framework that combines modality adapters, a shared encoder, presence-aware fusion, and self-supervised cross-view objectives to map five biological views into one latent space.
RepGene is trained self-supervised using frozen foundation-model features and without access to downstream benchmark labels. The learned representations are frozen and assessed using benchmark-defined linear probes under strict train–test splits.
□ Fast genomic read alignment with minibwa
https://arxiv.org/abs/2606.15357
minibwa, the next iteration of BWA-MEM. Minibwa aligns standard WGS short reads and Hi-C reads like BWA-MEM and additionally maps accurate long reads like minimap2 and it natively supports BS-seq data.
Minibwa finds properly paired reads using the same logic as BWA-MEM. If a read is mapped but its mate is not mapped nearby, minibwa locally rescues the mate in a window close to the mapped read.
□ MGA: a tool for haplotype-mixed assembly of long and accurate reads
https://link.springer.com/article/10.1186/s13059-026-04128-5
The Mosaic Genome Assembler (MGA) algorithm extends the set of operations used in block generation algorithms. Another improvement is the use of the multiplex de Bruijn graph, which increases assembly contiguity and minimizes phase-switches.
MGA iteratively performs operations of detouring, dewhirling, decoupling, repairing broken tips, and contracting short edges. MGA connects contigs that were separated by the drops in coverage using either their short overlaps or reads spanning these contigs.
MGA generates an assembly graph with increased contiguity. MGA identifies cognate contigs—those largely contained within another contig—and removes (deduplicates) them to avoid redundancy.
□ Praxis-BGM: Clustering of Omics Data Using Semi-Supervised Transfer Learning for Gaussian Mixture Models via Natural-Gradient Variational Inference
https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btag395/8709948
Praxis-BGM, a natural-gradient variational inference framework for Gaussian mixture models that incorporates informative priors—cluster-specific means, covariances, and structural connectivity—from large-scale reference data with robust cluster structures.
Praxis-BGM enables semi-supervised transfer learning on a small target dataset. Using the Variational Online Newton (VON) algorithm, it derives natural-gradient updates for the standard parameters of GMMs.
□ Paradoxical gene regulation explained by competition for genomic sites
https://www.biorxiv.org/content/10.1101/2025.11.27.691022v2
Competition between an activator and a repressor can produce counterintuitive outcomes in which increasing the abundance of a nominally positive regulator reduces target-gene expression. This effect does not reflect a change in the intrinsic biochemical roles of either regulator.
Instead, it arises from redistribution of regulators across a common pool of decoy sites, a form of retroactivity in which downstream binding interactions feed back onto regulator availability and reshape effective input signals.
The mechanistic reaction-network model developed here predicted that activator induction could release repressors from decoy sites and thereby enhance repression at the promoter, a prediction that was validated experimentally using a minimal CRISPRa/CRISPRi circuit.
Competitive binding at decoy sites was implemented through overlapping binding sequences, and systematic perturbations confirmed that exclusive binding and target affinity are essential determinants of the paradoxical regime.
□ ExpressionVAE: Elucidating the Design Space of Generative Models for Single-Cell Perturbation Prediction
https://www.biorxiv.org/content/10.64898/2026.06.15.732063v1
ExpressionVAE (eVAE), a discrete-latent perturbation model that compresses each cell into a short sequence of discrete codes through a finite-scalar-quantization (FSQ) bottleneck and trains a perturbation-conditioned discrete prior over those codes.
On Replogle and Parse 1M, eVAE sets a new state of the art on every distributional metric and leads on most cell-eval perturbation metrics, with Fréchet distance and MMD2 roughly 3 to 20x lower than the strongest continuous-latent baseline.
□ Movi 2: Fast and Space-Efficient Queries on Pangenome
https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btag362/8710943
Movi 2 greatly reduces size and memory footprint of move structure-based indexes. The most compressed version of Movi 2 reduces the Movi index’s space footprint more than fivefold.
Movi 2 achieves advantageous time and space tradeoffs when applied to large pangenome collections, including both the first and second releases of the Human Pangenome Reference Consortium (HPRC) collection, the latter of which spans over 460 human haplotypes.
□ A causal reinforcement learning framework for reliable gene regulatory network inference
https://link.springer.com/article/10.1186/s12859-026-06511-2
A state-action mapping and an experience replay mechanism are defined to represent the actions and states of GRN inference under the reinforcement learning strategy, and these elements are stored.
Subsequently, a deep Q-network (DQN) composed of k-nearest neighbors (KNN), a structural causal model (SCM) layer, and a variational graph autoencoder (VGAE) is constructed as the core inference architecture to generate corresponding actions.
Action rewards are calculated using the BIC combined with the network loss function. This guides the model to favor causal networks with simplicity in structure and strong biological interpretability, and the inference results are ultimately output after iterative optimization.
□ GOATEA: gene set enrichment analysis in R with shiny interactive visualizations
https://link.springer.com/article/10.1186/s12859-026-06541-w
Geneset Ordinal Association Test Enrichment Analysis (GOATEA) extends the GOAT algorithm with interactive visualization, multi-contrast comparison.
GOATEA integrates gene-and network-based context for bottom-up pathway analysis, enabling compreensive enrichment analysis.
□ GenOT: generative optimal transport enables spatiotemporal interpolation and generation in cross-platform spatial transcriptomics
https://link.springer.com/article/10.1186/s13059-026-04166-z
GenOT fuses gene expression and spatial information through a multi-scale graph convolutional network with an encoder-shared decoder structure, combined with contrastive learning to enhance the quality of gene expression embeddings.
GenOT designed a generative strategy based on OT barycenter technology, which enables spatiotemporal imputation of spatial transcriptomics data across slices/platforms.
GenOT follows an "encode-generate" paradigm, excelling not only in traditional spatial domain identification tasks but also demonstrating cross-platform and cross-slice generative capabilities.
□ Finding stable clusterings of single-cell RNA-seq data
https://www.biorxiv.org/content/10.1101/2025.09.17.672302v5
Constructing a pipeline that takes a UMI count matrix as input and produces clusterings of a range of sizes. Clusterings are generated for samples of cells. Comparing the clusterings of the samples with the same-size clusterings of the full set of cells gives stability estimates.
Clusterings are compared using what Meili˘a calls the misclassification error distance (MED). Von Luxburg calls it the minimal matching distance. Its distribution across samples characterizes a clustering’s stability.
For each cluster and sample a membership-based cluster misclassification error rate (CMER) is defined. Its distribution across samples characterizes the cluster’s stability.
□ GENATATORs: ab initio Gene Annotation With DNA Language Models
https://www.biorxiv.org/content/10.64898/2026.06.17.732686v1
GENATATORs are a family of fine-tuned DNA language model–based gene annotation models for accurate ab initio gene annotation from DNA sequence alone.
GENATATORs combine long-context modeling, multispecies training, and biologically informed post-processing to achieve state-of-the-art reconstruction of coding and non-coding gene structures, including UTRs and lncRNAs.
□ AutoZyme: An Autonomous Agentic Framework to Optimize Bioinformatics Software
https://www.biorxiv.org/content/10.64898/2026.06.12.731250v1
AutoZyme, an autonomous agentic framework for scientific software optimization. AutoZyme consists of five main agents: task setup, benchmark initialization, optimization, validation, and packaging.
AutoZyme also incorporates an Auditor agent that guards against agent hacking both after benchmark initialization and after optimization.
□ multiRF: Multivariate Random Forests for Cross-Modal Multi-Omics Integration
https://www.biorxiv.org/content/10.64898/2026.06.17.732933v1
MULTIRF learns a sample-level cross-omics neighborhood graph, which can capture nonlinear dependence, keep local sample weights, and be split into shared and modality-specific similarities without explicit latent-factor estimation or rank assumptions.
The learned cross-omics weights estimate the portion of each omics layer predictable from the others, and the remaining residual is treated as modality-specific signal.
A response subsampling step controls the multivariate split criterion for high-dimensional blocks, so the weight matrix is less dominated by high-dimensional noise.
□ VCBench: A Multi-Dimensional Benchmark for Single-Cell Foundation Models
https://www.biorxiv.org/content/10.64898/2026.06.18.733146v1
Single-cell foundation models are increasingly positioned as virtual cells, yet their capabilities are assessed by fragmented, largely single-task benchmarks that obscure where these models improve on simple baselines.
VCBench addresses this by synthesizing four independent virtual-cell frameworks into seven capability dimensions: perturbation response prediction, cross-species universality, GRN inference, modality integration, temporal dynamics, multi-scale integration, and in silico experimentation.
They evaluate five foundation models (Geneformer, scGPT, UCE, TranscriptFormer, Arc State) against pre-registered linear and nearest-neighbor baselines across the five testable dimensions, and report three findings.
□ fastQpick: scalable bootstrap and subsampling of FASTQ reads
https://www.biorxiv.org/content/10.64898/2026.06.23.734068v1
fastQpick enables fast, memory-efficient sampling of DNA-seq or RNA-seq FASTQ data with replacement. It is useful for generating bootstrap replicates to estimate technical variance in downstream analyses and for subsampling large datasets for testing and benchmarking.
fastQpick efficiently processes large libraries by streaming files in two passes by default: first to count reads and create a hash-based counter, then to write the sampled reads. It generates a full-size bootstrap replicate in under 30 minutes for a 500-million-read library.
□ A-liner: linear alignment visualizer for genome comparisons
https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btag408/8716316
A-liner, a command-line tool for linear visualization of genome-scale sequence alignments. A-liner is designed to complement existing visualization tools by focusing on the generation of integrated, publication-ready figures within a reproducible workflow.
A-liner enables the simultaneous visualization of sequence alignments together with gene annotations, highlighted genomic regions, coordinate scales, and user-defined quantitative tracks.
□ Bamsnap-LRS: an automated batch visualization tool for long-read sequencing alignments
https://www.biorxiv.org/content/10.64898/2026.06.21.733121v1
Bamsnap-LRS, an automated CLI tool for high-throughput LRS alignment visualization. It supports long-read-specific features, phased SNP inspection, and publication-ready batch figure generation within a unified framework for genomic, transcriptomic, and haplotype-aware analyses.
Bamsnap-LRS provides three visualization modes: DNA, RNA, and Highlight. DNA mode displays standard long-read alignments, coverage, and SV evidence within genomic regions.
RNA mode visualizes spliced alignments and transcript structures from long-read transcriptomic data. Highlight mode integrates SNPs from the input VCF with LRS alignment to facilitate the inspection of SNP linkage and haplotype consistency.
□ PerturbPlan: An analytical framework for designing Perturb-seq experiments
https://www.biorxiv.org/content/10.64898/2026.05.22.727199v1
PerturbPlan, an analytical framework and interactive web application (www.perturbplan.com) for designing Perturb-seq and TAP-seq experiments.
PerturbPlan resolves the core technical problem of power calculation by developing a novel analytical formula approximating the power of sceptre, 2,2 a state-of-the-art tool for testing perturbation-expression association in single-cell CRISPR screens.
This formula overcomes the computational limitations of simulation-based power calculation, enabling millisecond-speed power calculation for each parameter configuration and, in turn, real-time experimental design recommendations.
□ pygenoscape: a Python package for spatial interpolation and visualization of genetic distance landscapes
https://academic.oup.com/bioinformaticsadvances/article/6/1/vbag173/8711925
pygenoscape transforms pairwise genetic distance data into continuous spatial representations of genetic turnover. It accepts precomputed genetic distance matrices or aligned nucleotide sequence data and combine distance embedding, geographic projection and spatial interpolation.
pygenoscape operates on pairwise distance matrices together with geographic coordinates and supports both externally
generated genomic distances and internally computed sequence-based distances from aligned FASTA files.
Pairwise genetic distances are transformed into a low-dimensional representation using principal coordinate analysis (PCoA), and the resulting coordinate axis is interpolated across geographic space using radial basis function interpolation.
