Data analysis, machine learning, LLM apps, and prompt design (9.1K skills)
Generate financial market forecasts with Kronos using time-series data, meaning observations recorded over time. An agent then adjusts the numerical forecast using recent news sentiment and reasoning, combining a quantitative baseline with a subjective assessment.
Analyze financial text with local FinBERT or an LLM to identify positive, negative, or neutral sentiment. Return a numeric score and short reason, with support for batch processing unanalyzed news using FinBERT.
Find mainland Chinese A-share, Hong Kong, and US stocks by company name or ticker. Retrieve historical prices and trading volume for a chosen period, plus company information such as sector, market capitalization, and price-to-earnings ratio.
Extract protein abundance estimates from mass spectrometry data for comparisons between samples. Covers label-free measurements, TMT/iTRAQ chemical tags, and SILAC metabolic labels, with normalization, protein-level summaries, and heavy-to-light ratio calculations.
Remove sequencing adapter sequences from FASTQ files with Cutadapt or Trimmomatic. Handle single-end and paired-end reads, select standard or custom adapters, filter trimmed reads, and check the resulting files with FastQC before alignment.
Build and manage spectral libraries, reference collections for identifying peptides in proteomics data. Use measured or predicted spectra, convert library formats, check quality, and merge libraries for DIA, a mass spectrometry acquisition method.
Analyze chemical modifications added to proteins after production, including phosphorylation, acetylation, and ubiquitination. Filter modification sites by confidence, examine surrounding sequence patterns, and compare site quantities while accounting for changes in overall protein abundance.
Generate FASTQ sequencing quality reports with FastQC and combine them with MultiQC. Inspect quality at each base position, adapter contamination, GC composition bias, and duplicate reads, then compare samples or check results before and after trimming.
Measure distances, angles, and contacts in protein structures using Bio.PDB. Compare structures through superposition and RMSD, a measure of positional differences, and calculate solvent accessible surface area to examine exposed and buried residues.
Coordinate bioinformatics work by identifying input files and selecting specialized skills. Plan analyses with several steps, combine results into Markdown reports, and collect commands, environment details, and data checksums so the analysis can be repeated.
Check proteomics data before further analysis by summarizing detected proteins, missing values, measurement distributions, and agreement between replicates. Generate quality plots and summary metrics, and investigate whether processing batches explain differences between samples.
Analyze gene lists against WikiPathways to find biological pathways with more matching genes than expected. Test ranked expression changes, compare gene groups, and create plots or CSV tables using R tools.
Compare protein abundance between experimental groups using statistical tests in R or Python. Adjust for testing many proteins, filter results by significance and change size, and visualize candidate proteins in a volcano plot.
Assign taxonomic names to ASVs, distinct DNA sequences obtained from amplicon processing. Use reference databases such as SILVA, GTDB, or UNITE with DADA2, QIIME2, or IDTAXA, set confidence thresholds, and organize results in phyloseq.
Read, write, and convert molecular files with RDKit and Open Babel. Prepare compound libraries for analysis by checking structures and standardizing chemical representations, including charges, salts, and alternative forms of the same molecule.
Import mass spectrometry files and protein quantification results into Python or R for analysis. Access spectra, organize protein measurements by sample, remove flagged contaminants and reverse hits, and assess missing values across proteins and samples.
Process paired FASTQ files containing reads from both ends of DNA fragments with Biopython. Keep R1/R2 records together while filtering by quality or length, check pair IDs and counts, and combine or split alternating records.
Use RDKit to turn molecular structures into numerical properties and fingerprints, which encode structural features. Prepare compound data for machine learning, compare molecular similarity, and filter compounds using drug-likeness criteria.
Find biological functions overrepresented in a gene list using clusterProfiler. Analyze Gene Ontology categories describing processes, molecular functions, and cellular locations, convert gene identifiers, choose background genes, and export results with redundant terms reduced.
Turn clusterProfiler enrichment results, which identify overrepresented biological functions or pathways, into figures with enrichplot in R. Create summary charts, gene relationship networks, and gene set enrichment plots, customize their appearance, and export PDF or PNG files.
Explore protein structure files with Biopython Bio.PDB. Access individual chains, residues, and atoms, extract amino acid sequences and atomic coordinates, filter structural components, and count them across models, including structures with multiple conformations.
Combine gene expression, methylation, and other molecular measurements into one patient similarity network using SNFtool in R. Discover patient groups, visualize their relationships, and compare the resulting clusters with known subtypes, individual datasets, or survival outcomes.
Prepare RNA, protein, and methylation datasets for joint analysis in R. Normalize each data type, correct differences between measurement batches, match samples and gene identifiers, handle missing values, and export consistently scaled datasets for integration tools.
Identify translated regions in Ribo-seq data, which records ribosome activity along RNA. Use RiboCode to detect known and novel open reading frames, then quantify their translation and compare counts across experimental conditions.