mims-harvard/tooluniverse/tooluniverse-systems-biology

Systems Biology & Pathway Analysis

Comprehensive pathway and systems biology analysis integrating multiple curated databases to provide multi-dimensional view of biological systems, pathway enrichment, and protein-pathway relationships.

When to Use This Skill

Triggers:

• "Analyze pathways for this gene list"
• "What pathways is [protein] involved in?"
• "Find pathways related to [keyword/process]"
• "Perform pathway enrichment analysis"
• "Map proteins to biological pathways"
• "Find computational models for [process]"
• "Systems biology analysis of [genes/proteins]"

Use Cases:

1. Gene Set Analysis: Identify enriched pathways from RNA-seq, proteomics, or screen results
2. Protein Function: Discover pathways and processes a protein participates in
3. Pathway Discovery: Find pathways related to diseases, processes, or phenotypes
4. Systems Integration: Connect genes → pathways → processes → diseases
5. Model Discovery: Find computational systems biology models (SBML)
6. Cross-Database Validation: Compare pathway annotations across multiple sources

COMPUTE, DON'T DESCRIBE

When analysis requires computation (statistics, data processing, scoring, enrichment), write and run Python code via Bash. Don't describe what you would do — execute it and report actual results. Use ToolUniverse tools to retrieve data, then Python (pandas, scipy, statsmodels, matplotlib) to analyze it.

Domain Reasoning: Enrichment vs Causation

Pathway analysis answers: which biological processes are enriched in my gene list? But enrichment is not causation. A pathway being enriched means your gene list overlaps it more than expected by chance. Ask: is the enrichment driven by a few hub genes, or by many genes distributed across the pathway? A pathway with 3 input genes but 200 annotated members is less informative than one where 15 of 40 members are in your list.

LOOK UP DON'T GUESS: pathway membership, gene-to-pathway assignments, and enrichment statistics. Do not assume a gene is in a pathway — use Reactome, KEGG, or Enrichr to verify. Pathway databases disagree on membership; cross-validate key findings across at least two sources.

Core Databases Integrated

Workflow Overview

Input → Phase 1: Enrichment → Phase 2: Protein Mapping → Phase 3: Keyword Search → Phase 4: Top Pathways → Report

Phase 1: Pathway Enrichment Analysis

When: Gene list provided (from experiments, screens, differentially expressed genes)

Objective: Identify biological pathways statistically over-represented in gene list

Tools & Workflow

1. Submit gene list to Enrichr/Reactome. 2. Sort by adjusted p-value < 0.05. 3. Report top 10-20 pathways with IDs, p-values, and overlapping genes. If no enrichment, note explicitly.

Phase 2: Protein-Pathway Mapping

When: Protein UniProt ID provided

Objective: Map protein to all known pathways it participates in

Tools Used

Reactome_map_uniprot_to_pathways:

• Input:
• uniprot_id: UniProt accession (e.g., "P53350")
• Output: Array of Reactome pathways containing this protein

Reactome_get_pathway_reactions:

• Input:
• stId: Reactome pathway stable ID (e.g., "R-HSA-73817")
• Output: Array of reactions and subpathways
• Use: Get mechanistic details of pathways

Workflow

1. Map UniProt ID to Reactome pathways
2. Get all pathways this protein appears in
3. For top pathway (or user-specified):

• Retrieve detailed reactions and subpathways
• Extract event names, types (Reaction vs Pathway)
• Note disease associations if present

Decision Logic

• Multiple pathways: Report all pathways, prioritize by hierarchical level
• Top pathway details: Get detailed reactions for 1-3 most relevant
• Versioned IDs: Reactome uses unversioned IDs - strip version if present
• Empty results: Check if protein ID valid; suggest alternative databases if Reactome empty

Phase 3: Keyword-Based Pathway Search

When: User provides keyword or biological process name

Objective: Search multiple pathway databases to find relevant pathways

Tools

Search all databases in parallel. Group results by pathway concept. BioModels often returns empty — this is normal.

Phase 4: Top-Level Pathway Catalog

When: Always included to provide context

Objective: Show major biological systems/pathways for organism

Tools Used

Reactome_list_top_pathways:

• Input: species (e.g., "Homo sapiens")
• Output: Array of top-level pathway categories
• Use: Provides hierarchical pathway organization

Workflow

1. Retrieve top-level pathways for specified organism
2. Display pathway categories (metabolism, signaling, disease, etc.)
3. Serve as reference for pathway hierarchy

Decision Logic

• Always show: Provides context even if other phases empty
• Organism-specific: Filter by species of interest
• Hierarchical view: These are parent pathways with many subpathways

Output Structure

Create a markdown report progressively: header → Phase 1 enrichment results → Phase 2 protein mapping → Phase 3 keyword search → Phase 4 top pathway catalog. Note empty results explicitly; never silently omit them. Include pathway IDs for follow-up.

Tool Parameter Reference

Critical Parameter Notes (from testing):

Response Format Notes:

• Reactome: Returns list directly (not wrapped in {status, data})
• Pathway Commons: Returns dict with total_hits and pathways
• Others: Standard {status: "success", data: [...]} format

Domain Reasoning: Enzyme Kinetics & Metabolic Analysis

LOOK UP DON'T GUESS: Km values, kcat values, cofactor requirements, and optimal pH/temperature for specific enzymes. Use BindingDB_search_by_target, ChEMBL_get_molecule, BRENDA_get_enzyme_info (requires BRENDA_EMAIL + BRENDA_PASSWORD env vars; free academic registration at brenda-enzymes.org) (if available), or EuropePMC_search_articles to retrieve published kinetic parameters. Do not estimate Km from first principles.

Michaelis-Menten Kinetics

The foundational model: v = Vmax * [S] / (Km + [S])

• Km = substrate concentration at half-maximal velocity. NOT binding affinity (Km = (koff + kcat) / kon).
• Vmax = maximum velocity = kcat * [E_total]. Proportional to enzyme concentration.
• kcat = turnover number = molecules of substrate converted per enzyme per second.
• Catalytic efficiency = kcat / Km. The "best" enzymes approach the diffusion limit (~10^8 M^-1 s^-1).

To determine Km and Vmax from data: use Lineweaver-Burk (1/v vs 1/[S]), Eadie-Hofstee (v vs v/[S]), or nonlinear regression (preferred — avoids distortion from reciprocal transforms). See enzyme_kinetics.py in skills/tooluniverse-computational-biophysics/scripts/.

Allosteric Regulation & Cooperative Binding

Not all enzymes follow Michaelis-Menten. Sigmoidal v-vs-[S] curves indicate cooperativity.

• Hill equation: v = Vmax * [S]^nH / (K0.5^nH + [S]^nH)
• Hill coefficient (nH): nH = 1 (no cooperativity), nH > 1 (positive, e.g., hemoglobin O2 binding nH ~ 2.8), nH < 1 (negative cooperativity).
• K0.5: substrate concentration at half-maximal velocity (analogous to Km but not identical for cooperative systems).
• Allosteric activators shift the curve LEFT (lower K0.5). Allosteric inhibitors shift it RIGHT (higher K0.5) or reduce Vmax.

Enzyme Inhibition Types

To determine Ki: measure v at multiple [I] and [S], fit to the appropriate model. The enzyme_kinetics.py script handles competitive, uncompetitive, and noncompetitive inhibition calculations.

Troubleshooting "No Activity" Results

When a purified enzyme shows no catalytic activity, systematically check:

1. Oligomeric state: Many enzymes are obligate dimers/tetramers. Dilute protein may dissociate. Check with SEC, native PAGE, or DLS. Concentrate sample or add stabilizing agents (glycerol, specific ions).
2. Cofactors: Metal ions (Zn2+, Mg2+, Mn2+), coenzymes (NAD+, FAD, PLP), or prosthetic groups may be lost during purification. LOOK UP the enzyme's cofactor requirements and supplement the assay buffer.
3. pH: Most enzymes have a sharp pH optimum. Even 1 pH unit off can reduce activity 10-fold. Buffer at the literature-reported optimal pH.
4. Temperature: Standard assays at 25C or 37C. Thermophilic enzymes need 50-80C. Psychrophilic enzymes denature above 30C.
5. Reducing environment: Many enzymes need DTT or beta-mercaptoethanol to maintain active-site cysteines in reduced form.
6. Substrate: Wrong isomer (D- vs L-), wrong oxidation state, or degraded substrate. Use fresh substrate and verify by a positive control enzyme.
7. Inhibitors in buffer: EDTA chelates essential metals. Phosphate competes at phospho-binding sites. Detergents can denature.
8. Protein folding: Inclusion body protein may be misfolded even after refolding. Check by CD spectroscopy or thermal shift assay.

Metabolic Flux Analysis Reasoning

Metabolic flux analysis (MFA) quantifies the rates of metabolic reactions in vivo, not just enzyme activities in vitro.