Positional Encoding for Spectral Sequences

Embeds sequential information into mass spectrum m/z-intensity pairs for transformer-based MS/MS interpretation in IDSL_MINT.

triggers:preparing mass spectrum input tensorstransformer encoder layers in IDSL_MINTpositional encoding for spectral sequences

GitHub SKILL.md

What it does

This skill implements positional encoding to embed ordinal and sequential information into mass spectrum data. This allows transformer self-attention layers to distinguish the position of peaks in a spectrum, which is critical for accurate MS/MS interpretation.

When to use it

Use this skill specifically when preparing mass spectrum input tensors for transformer encoder layers within the IDSL_MINT framework. It should be applied after vectorizing m/z and intensity features but before the multi-head self-attention blocks.

What's included

Instructions: Precise mathematical formulation for sinusoidal position encodings (PE) following the 'Attention is All You Need' paradigm.
Implementation: A PyTorch-based example for computing and broadcasting the encoding matrix across batches.
Evaluation: Specific signals for validating tensor shapes, numerical stability (no NaNs), and sinusoidal patterns.

Compatible agents

Scientific AI agents equipped with PyTorch and RDKit, specifically those targeted at bioinformatics, cheminformatics, and metabolomics research.

Not yet audited

This skill has not been reviewed by our automated audit pipeline yet.

Information

Repository: asb-skill-collections
Stars: 14

More from asb-skill-collections

GitHub Actions Workflow Inspector

Verify and locally reproduce GitHub Actions CI workflows to validate badge status and build integrity.

Injection Sequence Annotation

Identifies and validates QCpool sample spacing in Sciex Multiquant metabolomics/lipidomics exports.

PyTorch Pretrained Weight Loader

Load pretrained PyTorch model checkpoints to rank candidate molecules or score spectra in metabolomics without retraining.

Related Skills

Seq Wrangler

Sequence data QC, trimming, alignment, and BAM processing pipeline helper — wraps FastQC, fastp/Trimmomatic, BWA/Bowtie2/Minimap2, and SAMtools for automated re

ToolUniverse Gateway

Unified agent gateway to 1000+ scientific tools (bioinformatics, drug discovery, genomics) via Harvard's ToolUniverse, returning JSON for downstream workflows.

MatchMS (mass spectrometry processing)

Mass spectrometry data processing: import/export mzML/MGF/MSP, filter and normalize spectra, compute spectral similarity (cosine/modified cosine), and build rep

matchms — Mass Spectrometry Processing

Python library skill for importing, filtering, and comparing mass spectra (mzML/MGF/MSP); builds reproducible preprocessing pipelines and computes spectral simi

MatchMS — Mass Spectrometry Utilities

Import, process and compare mass spectrometry spectra (mzML/MGF/MSP) with filters, similarity metrics and reproducible pipelines for metabolomics workflows.

KEGG Pathway Enrichment (clusterProfiler)

Run KEGG pathway and module enrichment using clusterProfiler (enrichKEGG/enrichMKEGG) to find metabolic and signaling pathways over-represented in a gene list.

Network Toxicology & Docking Research Planner

Generates four-tier network toxicology + molecular docking study designs (Lite → Publication+) with step-by-step workflows, validation strategies, figure plans,

Bizard — Biomedical Visualization Atlas

A comprehensive atlas of biomedical data visualization tutorials and examples (R/Python/Julia) that provides reproducible code and guidance for plots used in bi

Positional Encoding for Spectral Sequences

What it does

When to use it

What's included

Compatible agents