science

academic-publication-workflow

A collection of skills for managing the end-to-end process of academic research publication, including topic selection, writing, review, and submission to specific journals.

These skills differ by their specific focus on either general academic writing and review processes or the detailed procedural requirements of specific conferences and journals like CIKM, Nature, and various American Medical Journals.

Help me select a novel topic for a CIKM paper and draft the initial abstract.
Review my manuscript for Nature standards and generate the supplementary materials for submission.

atac-seq-analysis

A suite of computational methods for processing, quality controlling, and interpreting ATAC-seq data to identify open chromatin regions and regulatory elements.

The skills differ by their specific analytical focus, ranging from basic preprocessing and peak calling to advanced applications like single-cell analysis, motif deviation, and enhancer-gene linking.

Call peaks and perform quality control on bulk ATAC-seq data to identify accessible genomic regions.
Analyze single-cell ATAC-seq data to detect differential accessibility and link enhancers to target genes.

bio-visualization

A collection of skills for creating, customizing, and interpreting various types of biological data visualizations, including plots, tracks, and interactive figures.

These skills focus specifically on the graphical representation and visualization of biological data, distinguishing them from data conversion, ontology mapping, or database evidence retrieval.

Create a volcano plot with customized color palettes to highlight differentially expressed genes.
Generate a genome browser track visualization to display specific genomic features.

alignment-processing

A suite of skills for managing, validating, and manipulating sequence alignment files in SAM/BAM formats.

These skills focus specifically on the structural manipulation, quality control, and format management of alignment data, distinct from higher-level biological interpretation or specific amplicon processing.

Sort and index a large BAM file to enable random access for downstream analysis.
Validate the header and structure of a SAM file to ensure it meets specification standards.

biomedical-research-and-analysis

A suite of tools for executing biomedical research workflows, querying biological databases, and analyzing clinical or genomic data.

These skills differ by their specific function, ranging from general research agents and workflow orchestration to specific data extraction, database querying, and API integration for biomedical information.

Query the Open Targets database to identify potential drug targets for a specific disease and summarize the evidence.
Extract vital signs from a set of unstructured clinical notes and format them into a structured JSON output.

cheminformatics-identifier-conversion

A set of skills for converting chemical identifiers between formats such as IUPAC names, SMILES, and molecular formulas, including structure validation and data preparation.

This skill focuses specifically on the interconversion and validation of chemical identifiers and structures, whereas other skills in the cluster handle broader tasks like property lookup, safety data, or general chemical analysis.

Convert the SMILES string 'CCO' to its IUPAC name and molecular formula.
Validate the provided SMILES string and generate the corresponding InChI key.

clinical-trial-operations

A set of skills for designing, simulating, searching, and reporting on clinical trials.

These skills focus specifically on the lifecycle of clinical trials, from design and simulation to search and reporting, rather than general clinical documentation or variant interpretation.

Design a randomized controlled trial protocol for a new hypertension drug.
Simulate the power of an adaptive trial design with interim analyses.

clip-seq-analysis

A family of skills for processing, analyzing, and interpreting CLIP-seq data to identify RNA-protein binding sites and motifs.

These skills differ by their specific stage in the CLIP-seq workflow, ranging from raw data preprocessing and alignment to peak calling, motif analysis, and specialized deep learning or fine-tuning applications.

Run a full CLIP-seq pipeline to call peaks and analyze motifs from raw fastq files.
Perform quality control and preprocessing on CLIP-seq data before peak calling.

crispr-screen-analysis

A suite of skills for designing, executing, and analyzing CRISPR-based functional genomic screens, including library design, quality control, hit calling, and batch correction.

These skills focus specifically on the end-to-end workflow of CRISPR screening (from library design to hit calling and correction) rather than general gene editing, off-target prediction, or base editing analysis.

Design a CRISPR library targeting the kinome and run a screen QC check on the resulting sequencing data.
Perform batch correction on a pooled CRISPR screen dataset and call significant hits using MAGeCK.

differential-analysis

A family of statistical methods used to identify significant differences in molecular features, such as gene expression, binding, or accessibility, between experimental conditions.

The skills differ by the specific type of biological data being analyzed, ranging from transcriptomic expression and splicing to epigenomic binding, accessibility, and microbiome abundance.

Identify differentially expressed genes between treated and control samples using DESeq2.
Perform differential binding analysis on ChIP-seq data to find regions with altered protein occupancy.

computational-drug-discovery

A suite of AI and bioinformatics tools for identifying, validating, and optimizing potential therapeutic compounds.

These skills focus on the in-silico design, discovery, and analysis of new drugs and their mechanisms, distinct from general medical information retrieval or clinical rehabilitation tools.

Identify novel small molecule inhibitors for a specific protein target using deep learning models.
Predict the drug-drug interactions and repurposing potential of a candidate compound for a new disease indication.

gene-regulatory-network-analysis

A suite of skills for inferring, analyzing, and visualizing gene regulatory networks and their dynamic changes across conditions or omics layers.

These skills focus specifically on the construction, inference, and comparative analysis of regulatory relationships (GRNs) and co-expression structures, rather than general enrichment, essentiality, or panel design.

Infer a gene regulatory network from single-cell RNA-seq data using SCENIC and identify key regulons.
Compare differential gene regulatory networks between tumor and normal tissues to identify condition-specific transcription factors.

causal-genomics-inference

A suite of statistical methods for inferring causal relationships, genetic architecture, and molecular mechanisms from genomic and multi-omics data.

These skills focus on establishing causal directionality, partitioning heritability, and identifying molecular mediators, distinguishing them from purely descriptive association studies or non-genomic data analysis tools.

Perform a Mendelian randomization analysis to determine if elevated LDL cholesterol causally increases the risk of coronary artery disease.
Use mediation analysis to decompose the genetic effect on blood pressure into direct and indirect paths mediated by gene expression levels.

genome-assembly-and-annotation

A comprehensive suite of skills for constructing, quality-controlling, and biologically annotating genomic sequences from raw sequencing data.

These skills differ by their specific stage in the genomic pipeline, ranging from initial assembly and polishing to specialized annotation tasks for different organism types and gene classes.

Assemble a high-quality genome from PacBio HiFi reads and annotate all protein-coding genes and non-coding RNAs.
Compare two bacterial genome assemblies to identify structural variations and transfer functional annotations from a reference strain.

genomic-data-processing

A set of skills for processing, quality controlling, and analyzing raw genomic sequencing data and reference resources.

These skills focus on the technical handling, quality assurance, and preparation of genomic sequence data and reference materials, distinct from downstream clinical interpretation or specific model training tasks.

Run quality control and imputation on a new batch of short-read sequencing data using the appropriate reference panels.
Set up a long-read sequencing pipeline to process raw nanopore data for variant calling.

hic-analysis

A suite of skills for processing, analyzing, and visualizing Hi-C genomic interaction data.

The skills differ by their specific function in the Hi-C workflow, ranging from data input/output and matrix operations to specialized analyses like loop calling, TAD detection, and differential comparison.

Run a full Hi-C pipeline to detect TADs and visualize the contact matrix.
Perform differential analysis on two Hi-C datasets to identify significant changes in chromatin interactions.

aistats-paper-lifecycle

A comprehensive set of skills for managing the end-to-end process of preparing, submitting, and revising a paper for the AISTATS conference.

These skills are specifically tailored to the AISTATS conference guidelines and workflow, distinguishing them from general journal club activities, cover art generation, or ASPLOS-specific experimental tasks.

Help me structure the related work section and ensure my experiments are reproducible for the AISTATS submission.
Draft a response to the reviewers' comments and prepare the camera-ready version of my AISTATS paper.

mathematical-modeling-and-verification

A suite of skills for designing, coding, and auditing mathematical models and proofs using formal methods and computational tools.

These skills differ by their specific role in the mathematical workflow, ranging from code generation and review to logical auditing and method explanation.

Generate MATLAB code for a linear algebra model and review it for numerical stability.
Audit the consistency of a mathematical proof and explain the final method used.

molecular-structure-prediction

A family of skills focused on predicting and analyzing the 3D structures of biological macromolecules using computational models and databases.

These skills differ from other medical or drug discovery agents by specifically focusing on structural biology and conformational analysis rather than clinical planning, vector design, or general drug screening.

Predict the tertiary structure of a novel protein sequence using AlphaFold.
Retrieve the predicted structure of a specific antibody from the AlphaFold database.

model-lifecycle-management

A collection of skills for validating, optimizing, interpreting, and deploying machine learning models across various frameworks and domains.

These skills differ by their specific stage in the model lifecycle, ranging from data curation and parameter optimization to interpretability analysis, framework-specific conversion, and final deployment or evaluation.

Validate a bio-machine-learning model and export its evidence for medical technology compliance.
Fine-tune a transformer model using Unsloth and visualize the training metrics with TensorBoard.

molecular-computation

A suite of tools for processing, analyzing, and simulating molecular structures and their physical properties.

These skills differ by their specific focus on data I/O and standardization, descriptor calculation, or advanced simulation and generative modeling tasks.

Standardize a set of SMILES strings and calculate their molecular weight and logP descriptors.
Run a molecular dynamics simulation using OpenMM to analyze the stability of a protein-ligand complex.

protein-structure-and-design

A suite of tools for predicting, retrieving, analyzing, and designing protein structures and sequences.

These skills differ by their specific function in the protein lifecycle, ranging from structure prediction and retrieval to sequence design and interaction analysis.

Predict the 3D structure of a novel protein sequence using Boltz and validate it with Protein QC.
Retrieve the PDB structure of a known protein and use Foldseek to find similar structures in the database.

proteomics-analysis

A comprehensive suite of skills for processing, quantifying, and interpreting mass spectrometry-based proteomics data.

The skills differ by their specific stage in the proteomics workflow, ranging from initial data import and spectral library creation to peptide identification, protein inference, quantification, and specialized analyses like PTM mapping or DIA processing.

Run a full DIA proteomics pipeline to identify peptides and quantify protein abundance across multiple samples.
Import raw proteomics data and perform quality control checks before differential abundance analysis.

rna-velocity-analysis

Estimates cell state transitions and trajectory directions from unspliced and spliced mRNA dynamics in single-cell RNA-seq data.

This skill focuses specifically on single-cell RNA velocity and trajectory inference using scVelo, whereas other skills in the cluster address bulk RNA-seq quantification, alignment, or general differential expression.

Run scVelo to infer the direction of cell differentiation in a single-cell dataset.
Calculate latent time and identify driver genes for RNA velocity in a stem cell population.

bio-sequence-manipulation

A set of skills for reading, filtering, transforming, and analyzing biological sequence data.

These skills focus on the direct manipulation, transformation, and basic statistical analysis of sequence data objects, distinct from high-level similarity searching or general file compression utilities.

Filter a FASTA file to keep only sequences longer than 500 base pairs and calculate their GC content.
Read a FASTQ file, reverse-complement the sequences, and write the results to a new file.

single-cell-omics-analysis

A comprehensive suite of computational methods for processing, analyzing, and interpreting high-dimensional single-cell sequencing data.

These skills collectively cover the entire lifecycle of single-cell data analysis, from raw data input and preprocessing to advanced downstream analyses like trajectory inference and communication modeling.

Run a full scRNA-seq pipeline including preprocessing, clustering, and cell type annotation on a new dataset.
Identify cell-cell communication networks and infer developmental trajectories from single-cell RNA sequencing data.

spatial-transcriptomics-analysis

A comprehensive suite of skills for processing, analyzing, and visualizing spatial transcriptomics data, including preprocessing, domain identification, deconvolution, and multi-omics integration.

These skills differ by their specific analytical focus, ranging from data input and preprocessing to advanced tasks like deconvolution, multi-omics integration, and specialized agent-based workflows.

Preprocess a 10x Visium dataset, identify spatial domains, and visualize the resulting tissue architecture.
Integrate spatial transcriptomics with proteomics data to analyze spatial communication between cell types.

structural-variant-analysis

Detects and characterizes large-scale genomic rearrangements such as deletions, insertions, and inversions from sequencing data.

This skill focuses on identifying genomic rearrangements from sequencing data, whereas other skills in the cluster focus on 3D protein structure prediction, manipulation, or validation.

Identify all structural variants larger than 50bp in this long-read sequencing dataset.
Call structural variants from short-read data using discordant pairs and split reads.

tao-model-training

A collection of skills for training and fine-tuning various computer vision models using the TAO toolkit.

These skills differ by the specific computer vision architecture or task (e.g., detection, segmentation, depth estimation) they are designed to train.

Train a BEVFusion model for autonomous driving perception.
Fine-tune a Grounding DINO model for open-vocabulary object detection.

variant-analysis

A family of skills for detecting, filtering, annotating, and interpreting genetic variants from sequencing data.

These skills differ by their specific stage in the variant analysis pipeline, ranging from initial variant calling and normalization to downstream annotation, prioritization, and clinical interpretation.

Call and filter variants from raw FASTQ files using GATK best practices.
Prioritize and interpret candidate variants for a clinical case using population databases and pathogenicity predictors.