agentleFS
dailycafidailycafi/metabolism-skills/skills/metabolomics-analysis/clinical-metabolomics/SKILL.md

bio-metabolomics-clinical-reporting

Skill7 starsChanged 8 months agoMITView on GitHub

Interpret clinical metabolomics results for inborn errors of metabolism (IEM) screening, newborn screening, and diagnostic reporting. Use when: user has clinical metabolite panels, needs IEM differential diagnosis, wants to analyze acylcarnitine profiles or amino acid panels, or calculate z-scores against reference ranges. Triggers: newborn screening, IEM, inborn error of metabolism, acylcarnitine, amino acid panel, organic acid analysis, clinical diagnosis, tandem MS screening, PKU, MCADD, maple syrup urine disease, clinical metabolomics, reference range, z-score.

  • Installs packages

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 8 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before using code patterns, verify installed versions match.

    line 12
  2. Distinctive:

    If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API …

    line 15
  3. Distinctive:

    Approach: Read instrument export, apply internal standard correction, compute analyte ratios, and flag abnormal patterns.

    line 91
  4. Distinctive:

    Reference ranges: Always use age-matched and method-specific reference ranges.

    line 388
  5. Distinctive:

    Second-tier testing: Never report a positive IEM diagnosis from first-tier screening alone.

    line 389
  6. Distinctive:

    Internal standards: Use stable isotope-labeled internal standards for every analyte class.

    line 390
  7. Distinctive:

    Creatinine normalization: Always normalize urine organic acids to creatinine.

    line 392
  8. Distinctive:

    Ratio markers: Use diagnostic ratios (Phe/Tyr, C3/C2, C5/C2, C8/C10) in addition to absolute values.

    line 393

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Version Compatibility

Reference examples tested with: pandas 2.2+, scipy 1.12+, numpy 1.26+, matplotlib 3.8+, seaborn 0.13+

Before using code patterns, verify installed versions match. If versions differ:

  • Python: pip show <package> then help(module.function) to check signatures

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

Clinical Metabolomics Reporting

"Screen newborn dried blood spots for inborn errors of metabolism" → Process amino acid and acylcarnitine panels from LC-MS/MS, calculate z-scores against age-matched reference ranges, flag abnormal results, and generate differential diagnoses.

  • Python: pandas for data processing, scipy for z-scores, matplotlib for metabolite panels
  • Analytical: ion-exchange chromatography, LC-MS/MS, GC-MS for organic acids

Installation

pip install pandas numpy scipy matplotlib seaborn openpyxl

Clinical Reference Ranges and Z-Score Calculation

Goal: Define age-matched reference ranges for newborn screening metabolites and compute z-scores for patient results.

Approach: Store reference ranges as structured data, compute z-scores using (value - mean) / SD, and flag results outside clinical thresholds.

import pandas as pd
import numpy as np
from scipy import stats

reference_ranges = {
    'phenylalanine':         {'mean': 55.0,  'sd': 15.0, 'unit': 'umol/L', 'upper_cutoff': 120.0, 'lower_cutoff': 10.0},
    'tyrosine':              {'mean': 100.0, 'sd': 40.0, 'unit': 'umol/L', 'upper_cutoff': 250.0, 'lower_cutoff': 15.0},
    'leucine_isoleucine':    {'mean': 130.0, 'sd': 35.0, 'unit': 'umol/L', 'upper_cutoff': 250.0, 'lower_cutoff': 30.0},
    'methionine':            {'mean': 22.0,  'sd': 8.0,  'unit': 'umol/L', 'upper_cutoff': 50.0,  'lower_cutoff': 5.0},
    'citrulline':            {'mean': 18.0,  'sd': 7.0,  'unit': 'umol/L', 'upper_cutoff': 45.0,  'lower_cutoff': 3.0},
    'valine':                {'mean': 120.0, 'sd': 30.0, 'unit': 'umol/L', 'upper_cutoff': 250.0, 'lower_cutoff': 30.0},
    'C0_carnitine':          {'mean': 30.0,  'sd': 10.0, 'unit': 'umol/L', 'upper_cutoff': 60.0,  'lower_cutoff': 8.0},
    'C3_propionylcarnitine': {'mean': 2.0,   'sd': 0.8,  'unit': 'umol/L', 'upper_cutoff': 5.0,   'lower_cutoff': 0.2},
    'C5_isovalerylcarnitine':{'mean': 0.15,  'sd': 0.06, 'unit': 'umol/L', 'upper_cutoff': 0.45,  'lower_cutoff': 0.01},
    'C8_octanoylcarnitine':  {'mean': 0.10,  'sd': 0.05, 'unit': 'umol/L', 'upper_cutoff': 0.30,  'lower_cutoff': 0.01},
    'C14_1_tetradecenoylcarnitine': {'mean': 0.12, 'sd': 0.06, 'unit': 'umol/L', 'upper_cutoff': 0.60, 'lower_cutoff': 0.01},
    'C16_palmitoylcarnitine':{'mean': 3.5,   'sd': 1.2,  'unit': 'umol/L', 'upper_cutoff': 7.0,   'lower_cutoff': 0.5},
}

ref_df = pd.DataFrame(reference_ranges).T

def compute_z_scores(patient_results, ref_df):
    z_scores = {}
    for analyte, value in patient_results.items():
        if analyte in ref_df.index:
            ref = ref_df.loc[analyte]
            z = (value - ref['mean']) / ref['sd']
            z_scores[analyte] = {
                'value': value,
                'unit': ref['unit'],
                'z_score': round(z, 2),
                'percentile': round(stats.norm.cdf(z) * 100, 1),
                'flag': 'HIGH' if value > ref['upper_cutoff'] else ('LOW' if value < ref['lower_cutoff'] else 'NORMAL'),
            }
    return pd.DataFrame(z_scores).T

patient = {
    'phenylalanine': 280.0,
    'tyrosine': 45.0,
    'leucine_isoleucine': 140.0,
    'methionine': 25.0,
    'citrulline': 15.0,
    'C0_carnitine': 32.0,
    'C3_propionylcarnitine': 1.8,
    'C5_isovalerylcarnitine': 0.12,
}

report = compute_z_scores(patient, ref_df)

Newborn Screening Amino Acid Panel

Goal: Process a batch of dried blood spot (DBS) amino acid results from the LC-MS/MS instrument and generate a screening report.

Approach: Read instrument export, apply internal standard correction, compute analyte ratios, and flag abnormal patterns.

import pandas as pd
import numpy as np

raw_data = pd.read_csv('nbs_amino_acids_batch.csv')

def normalize_by_istd(raw_data, analyte_col, istd_col):
    return raw_data.assign(
        **{f'{analyte_col}_corrected': raw_data[analyte_col] / raw_data[istd_col]}
    )

istd_pairs = {'phenylalanine': 'phe_d5', 'tyrosine': 'tyr_d4',
              'leucine_isoleucine': 'leu_d3', 'methionine': 'met_d3',
              'citrulline': 'cit_d2', 'valine': 'val_d8'}

corrected = raw_data.copy()
for analyte, istd in istd_pairs.items():
    if analyte in corrected.columns and istd in corrected.columns:
        corrected[f'{analyte}_corrected'] = corrected[analyte] / corrected[istd]

corrected['phe_tyr_ratio'] = corrected['phenylalanine'] / corrected['tyrosine'].replace(0, np.nan)

screening_flags = pd.DataFrame({
    'sample_id': corrected['sample_id'],
    'phe_flag': corrected['phenylalanine'].apply(lambda x: 'ABNORMAL' if x > 120 else 'NORMAL'),
    'phe_tyr_ratio_flag': corrected['phe_tyr_ratio'].apply(lambda x: 'ABNORMAL' if x > 3.0 else 'NORMAL'),
    'cit_flag': corrected['citrulline'].apply(lambda x: 'HIGH' if x > 45 else ('LOW' if x < 3 else 'NORMAL')),
})

Acylcarnitine Profiling

Goal: Analyze acylcarnitine profiles from LC-MS/MS for fatty acid oxidation disorders and organic acidemias.

Approach: Compute diagnostic ratios (C3/C2, C5/C2, C8/C10), classify acylcarnitine patterns, and match against known disorder profiles.

import pandas as pd
import numpy as np

acylcarnitine_data = pd.read_csv('acylcarnitine_panel.csv', index_col='sample_id')

diagnostic_ratios = pd.DataFrame(index=acylcarnitine_data.index)
diagnostic_ratios['C3_C2'] = acylcarnitine_data['C3'] / acylcarnitine_data['C2'].replace(0, np.nan)
diagnostic_ratios['C5_C2'] = acylcarnitine_data['C5'] / acylcarnitine_data['C2'].replace(0, np.nan)
diagnostic_ratios['C5_C3'] = acylcarnitine_data['C5'] / acylcarnitine_data['C3'].replace(0, np.nan)
diagnostic_ratios['C8_C10'] = acylcarnitine_data['C8'] / acylcarnitine_data['C10'].replace(0, np.nan)
diagnostic_ratios['C14_1_C16'] = acylcarnitine_data['C14_1'] / acylcarnitine_data['C16'].replace(0, np.nan)
diagnostic_ratios['C0_C16_C18'] = acylcarnitine_data['C0'] / (
    acylcarnitine_data['C16'] + acylcarnitine_data['C18']
).replace(0, np.nan)

disorder_profiles = {
    'MCADD': {'C8': ('>',  0.30), 'C8_C10': ('>', 1.5), 'C6': ('>', 0.15)},
    'VLCADD': {'C14_1': ('>', 0.60), 'C14_1_C16': ('>', 0.1), 'C14': ('>', 0.70)},
    'PA': {'C3': ('>', 5.0), 'C3_C2': ('>', 0.25)},
    'MMA': {'C3': ('>', 5.0), 'C3_C2': ('>', 0.25)},
    'IVA': {'C5': ('>', 0.45), 'C5_C2': ('>', 0.03)},
    'GA1': {'C5DC': ('>', 0.15)},
    'CPT2': {'C16': ('>', 7.0), 'C18': ('>', 2.5), 'C0': ('<', 8.0)},
}

def screen_for_disorders(sample_values, ratios, profiles):
    matches = []
    combined = {**sample_values, **ratios}
    for disorder, criteria in profiles.items():
        criteria_met = 0
        total_criteria = len(criteria)
        for analyte, (op, threshold) in criteria.items():
            if analyte in combined:
                val = combined[analyte]
                if op == '>' and val > threshold:
                    criteria_met += 1
                elif op == '<' and val < threshold:
                    criteria_met += 1
        if criteria_met == total_criteria:
            matches.append({'disorder': disorder, 'confidence': 'HIGH'})
        elif criteria_met >= total_criteria * 0.5:
            matches.append({'disorder': disorder, 'confidence': 'POSSIBLE'})
    return matches

Organic Acid Analysis (GC-MS Urine)

Goal: Process urine organic acid results from GC-MS and flag diagnostic markers for IEM.

Approach: Normalize to creatinine, compute z-scores, and identify patterns characteristic of specific organic acidurias.

import pandas as pd
import numpy as np

organic_acids = pd.read_csv('urine_organic_acids.csv', index_col='sample_id')

creatinine = organic_acids.pop('creatinine_mmol_L')
oa_normalized = organic_acids.div(creatinine, axis=0)

# Reference ranges in mmol/mol creatinine (creatinine-normalized urine values)
oa_reference = {
    'methylmalonic_acid':      {'mean': 2.5, 'sd': 1.5,  'alert_threshold': 15.0},
    'methylcitric_acid':       {'mean': 1.0, 'sd': 0.8,  'alert_threshold': 5.0},
    '3_hydroxypropionic_acid': {'mean': 3.0, 'sd': 2.0,  'alert_threshold': 20.0},
    'glutaric_acid':           {'mean': 1.5, 'sd': 1.0,  'alert_threshold': 10.0},
    '3_hydroxyglutaric_acid':  {'mean': 2.0, 'sd': 1.2,  'alert_threshold': 8.0},
    'isovalerylglycine':       {'mean': 0.3, 'sd': 0.2,  'alert_threshold': 2.0},
    'suberylglycine':          {'mean': 0.1, 'sd': 0.08, 'alert_threshold': 1.0},
    'hexanoylglycine':         {'mean': 0.1, 'sd': 0.07, 'alert_threshold': 0.8},
    'orotic_acid':             {'mean': 1.0, 'sd': 0.6,  'alert_threshold': 5.0},
}

oa_ref_df = pd.DataFrame(oa_reference).T

def flag_organic_acids(sample_normalized, oa_ref_df):
    flags = {}
    for acid in sample_normalized.index:
        if acid in oa_ref_df.index:
            ref = oa_ref_df.loc[acid]
            val = sample_normalized[acid]
            z = (val - ref['mean']) / ref['sd']
            flags[acid] = {
                'value_per_creatinine': round(val, 2),
                'z_score': round(z, 2),
                'flag': 'CRITICAL' if val > ref['alert_threshold'] else ('HIGH' if z > 3 else 'NORMAL'),
            }
    return pd.DataFrame(flags).T

Diagnostic Algorithm: Pattern Matching

Goal: Map flagged metabolite patterns to specific IEM diagnoses with differential diagnoses.

Approach: Encode diagnostic rules as structured decision trees and return ranked differentials with confirmatory test recommendations.

import pandas as pd

diagnostic_rules = [
    {'condition': 'PKU', 'omim': 261600,
     'primary_markers': {'phenylalanine': ('>', 120)},
     'secondary_markers': {'phe_tyr_ratio': ('>', 3.0), 'tyrosine': ('<', 50)},
     'confirmatory': ['PAH gene sequencing', 'BH4 loading test']},
    {'condition': 'MSUD', 'omim': 248600,
     'primary_markers': {'leucine_isoleucine': ('>', 200)},
     'secondary_markers': {'valine': ('>', 250), 'alloisoleucine': ('>', 5)},
     'confirmatory': ['Alloisoleucine quantification', 'BCKDH enzyme assay']},
    {'condition': 'MMA', 'omim': 251000,
     'primary_markers': {'C3_propionylcarnitine': ('>', 5.0)},
     'secondary_markers': {'C3_C2': ('>', 0.25), 'methylmalonic_acid': ('>', 15)},
     'confirmatory': ['Urine organic acids', 'MUT/MMAA/MMAB gene panel', 'Vitamin B12 trial']},
    {'condition': 'IVA', 'omim': 243500,
     'primary_markers': {'C5_isovalerylcarnitine': ('>', 0.45)},
     'secondary_markers': {'isovalerylglycine': ('>', 2.0)},
     'confirmatory': ['Urine isovalerylglycine', 'IVD gene sequencing']},
    {'condition': 'MCADD', 'omim': 201450,
     'primary_markers': {'C8_octanoylcarnitine': ('>', 0.30)},
     'secondary_markers': {'C8_C10': ('>', 1.5), 'hexanoylglycine': ('>', 0.8)},
     'confirmatory': ['ACADM gene sequencing (c.985A>G common)', 'Urine acylglycines']},
    {'condition': 'Citrullinemia Type I', 'omim': 215700,
     'primary_markers': {'citrulline': ('>', 45)},
     'secondary_markers': {'argininosuccinate': ('<', 2.0)},
     'confirmatory': ['ASS1 gene sequencing', 'Plasma ammonia']},
]

def evaluate_diagnosis(patient_values, rules):
    results = []
    for rule in rules:
        primary_match = 0
        primary_total = len(rule['primary_markers'])
        for marker, (op, threshold) in rule['primary_markers'].items():
            if marker in patient_values:
                val = patient_values[marker]
                if (op == '>' and val > threshold) or (op == '<' and val < threshold):
                    primary_match += 1

        secondary_match = 0
        secondary_total = len(rule['secondary_markers'])
        for marker, (op, threshold) in rule['secondary_markers'].items():
            if marker in patient_values:
                val = patient_values[marker]
                if (op == '>' and val > threshold) or (op == '<' and val < threshold):
                    secondary_match += 1

        if primary_match == primary_total:
            confidence = 'LIKELY' if secondary_match > 0 else 'POSSIBLE'
            results.append({
                'condition': rule['condition'],
                'omim': rule['omim'],
                'confidence': confidence,
                'primary_markers_met': f'{primary_match}/{primary_total}',
                'secondary_markers_met': f'{secondary_match}/{secondary_total}',
                'confirmatory_tests': '; '.join(rule['confirmatory']),
            })
    return pd.DataFrame(results)

patient_values = {
    'phenylalanine': 280.0,
    'tyrosine': 45.0,
    'phe_tyr_ratio': 6.2,
    'leucine_isoleucine': 140.0,
    'C3_propionylcarnitine': 1.8,
    'C5_isovalerylcarnitine': 0.12,
    'C8_octanoylcarnitine': 0.08,
    'citrulline': 15.0,
}

diagnosis_report = evaluate_diagnosis(patient_values, diagnostic_rules)

Quality Control: Internal Standards and External QC Programs

Goal: Monitor analytical quality using internal standard recovery, external QC program results (ERNDIM, CDC NSQAP), and Levey-Jennings tracking.

Approach: Compute IS recovery rates, compare to external QC target values, and generate control charts.

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

batch_istd = pd.read_csv('batch_istd_areas.csv', index_col='sample_id')
median_istd = batch_istd.median()
recovery = batch_istd.div(median_istd) * 100

istd_pass = (recovery > 70) & (recovery < 130)
failed_samples = recovery.index[~istd_pass.all(axis=1)].tolist()

nsqap_results = pd.DataFrame({
    'analyte': ['phenylalanine', 'C3', 'C8', 'C5', 'citrulline'],
    'lab_value': [62.0, 2.1, 0.12, 0.16, 19.5],
    'peer_mean': [61.5, 2.05, 0.115, 0.155, 19.8],
    'peer_sd': [5.0, 0.3, 0.02, 0.025, 2.5],
})
nsqap_results['z_score'] = (nsqap_results['lab_value'] - nsqap_results['peer_mean']) / nsqap_results['peer_sd']
nsqap_results['acceptable'] = nsqap_results['z_score'].abs() < 2.0

def levey_jennings_chart(qc_values, analyte_name, target, sd):
    runs = np.arange(1, len(qc_values) + 1)
    fig, ax = plt.subplots(figsize=(12, 5))
    ax.plot(runs, qc_values, 'ko-', markersize=5)
    ax.axhline(y=target, color='green', linestyle='-', label='Target')
    for mult, color in [(2, 'orange'), (3, 'red')]:
        ax.axhline(y=target + mult * sd, color=color, linestyle='--')
        ax.axhline(y=target - mult * sd, color=color, linestyle='--')
    violations = [i for i, v in enumerate(qc_values) if abs(v - target) > 3 * sd]
    if violations:
        ax.plot([runs[i] for i in violations], [qc_values[i] for i in violations],
                'rv', markersize=10, label='Out of control')
    ax.set_xlabel('Run Number')
    ax.set_ylabel(f'{analyte_name} Concentration')
    ax.set_title(f'Levey-Jennings: {analyte_name}')
    ax.legend(loc='upper right')
    plt.tight_layout()
    plt.savefig(f'qc_chart_{analyte_name}.png', dpi=150)
    return fig

qc_phe = [58, 62, 55, 61, 59, 63, 57, 60, 64, 56, 61, 72, 59, 60, 58]
levey_jennings_chart(qc_phe, 'Phenylalanine', target=60.0, sd=5.0)

Metabolite Panel Visualization

Goal: Generate a clinical-style metabolite panel plot showing patient values against reference ranges.

Approach: Plot each analyte as a point with reference range bars, color-coded by flag status.

import matplotlib.pyplot as plt

def plot_metabolite_panel(report_df, title='Metabolite Screening Panel'):
    analytes = report_df.index.tolist()
    n = len(analytes)
    fig, ax = plt.subplots(figsize=(10, max(6, n * 0.4)))
    colors = {'NORMAL': '#4CAF50', 'HIGH': '#F44336', 'LOW': '#FF9800', 'CRITICAL': '#9C27B0'}
    for i, analyte in enumerate(analytes):
        row = report_df.loc[analyte]
        z, flag = row['z_score'], row['flag']
        ax.barh(i, z, color=colors.get(flag, '#757575'), height=0.6, alpha=0.7)
        ax.text(z + (0.1 if z >= 0 else -0.1), i, f"{row['value']} {row['unit']}",
                va='center', ha='left' if z >= 0 else 'right', fontsize=8)
    ax.set_yticks(range(n))
    ax.set_yticklabels(analytes, fontsize=9)
    ax.axvline(x=0, color='black', linewidth=0.8)
    for threshold, color in [(2, 'orange'), (-2, 'orange'), (3, 'red'), (-3, 'red')]:
        ax.axvline(x=threshold, color=color, linewidth=0.8, linestyle='--', alpha=0.5)
    ax.set_xlabel('Z-Score')
    ax.set_title(title)
    plt.tight_layout()
    plt.savefig('metabolite_panel.png', dpi=200)
    return fig

plot_metabolite_panel(report, title='Newborn Screening Panel - Patient 2024-001')

Best Practices

  • Reference ranges: Always use age-matched and method-specific reference ranges; newborn, pediatric, and adult cutoffs differ substantially for amino acids and acylcarnitines.
  • Second-tier testing: Never report a positive IEM diagnosis from first-tier screening alone; always recommend confirmatory testing (enzyme assay, molecular genetics, repeat specimen).
  • Internal standards: Use stable isotope-labeled internal standards for every analyte class; recovery should be 70-130% across the batch.
  • External QC: Participate in proficiency testing programs (ERNDIM for Europe, CDC NSQAP for US) and track z-scores longitudinally.
  • Creatinine normalization: Always normalize urine organic acids to creatinine; spot urine concentration varies by hydration status.
  • Ratio markers: Use diagnostic ratios (Phe/Tyr, C3/C2, C5/C2, C8/C10) in addition to absolute values; ratios improve specificity over single-analyte cutoffs.
  • Westgard rules: Apply multi-rule QC (1-3s, 2-2s, R-4s) rather than single-rule rejection to minimize false rejection of valid analytical runs.
  • Clinical context: Metabolite elevations can be transient (prematurity, TPN, maternal medication); always correlate with clinical presentation before reporting.

Related Skills

  • targeted-analysis - Standard curves and absolute quantification
  • normalization-qc - Batch QC and normalization methods
  • statistical-analysis - Group comparisons and biomarker discovery
  • pathway-mapping - Map abnormal metabolites to pathway context

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.