Bio Epidemiological Genomics Amr Surveillance Agent Skill

# Install AMRFinderPlus
conda install -c bioconda ncbi-amrfinderplus

# Update database
amrfinder -u

# Basic AMR detection from genome
amrfinder -n genome.fasta -o results.tsv

# With protein input (faster, more sensitive)
amrfinder -p proteins.faa -o results.tsv

# Specify organism for point mutations
amrfinder -n genome.fasta --organism Salmonella -o results.tsv

# Available organisms: Acinetobacter_baumannii, Campylobacter,
# Clostridioides_difficile, Enterococcus_faecalis, Enterococcus_faecium,
# Escherichia, Klebsiella, Neisseria, Pseudomonas_aeruginosa,
# Salmonella, Staphylococcus_aureus, Staphylococcus_pseudintermedius,
# Streptococcus_agalactiae, Streptococcus_pneumoniae, Streptococcus_pyogenes,
# Vibrio_cholerae

import pandas as pd

def parse_amrfinder(results_file):
    '''Parse AMRFinderPlus output

    Key columns:
    - Gene symbol: AMR gene name
    - Sequence name: Contig/protein where found
    - Element type: AMR, STRESS, VIRULENCE
    - Element subtype: AMR mechanism
    - Class: Drug class affected
    - Subclass: Specific drug affected
    - % Coverage: Alignment coverage (>90% typical cutoff)
    - % Identity: Sequence identity (>90% typical cutoff)
    '''
    df = pd.read_csv(results_file, sep='\t')

    # Filter high-confidence hits
    df = df[(df['% Coverage of reference sequence'] >= 90) &
            (df['% Identity to reference sequence'] >= 90)]

    return df


def summarize_amr_profile(results_df):
    '''Summarize AMR profile by drug class'''
    amr_only = results_df[results_df['Element type'] == 'AMR']

    summary = {
        'total_genes': len(amr_only),
        'drug_classes': amr_only['Class'].nunique(),
        'by_class': amr_only.groupby('Class')['Gene symbol'].apply(list).to_dict()
    }

    return summary

# ResFinder for acquired resistance genes
# Web: https://cge.cbs.dtu.dk/services/ResFinder/

# Command line via KMA
kma -i reads_1.fq reads_2.fq -o output -t_db resfinder_db -1t1

# Or use CGE Docker
docker run --rm -v $(pwd):/data cgetools/resfinder \
    -i /data/genome.fasta -o /data/results -db_res /db/resfinder_db

def analyze_amr_trends(samples_df, date_col='collection_date', gene_col='Gene symbol'):
    '''Analyze AMR gene prevalence over time

    For surveillance programs tracking:
    - Emergence of new resistance
    - Increasing prevalence of known resistance
    - Geographic spread patterns
    '''
    # Group by time period
    samples_df['period'] = pd.to_datetime(samples_df[date_col]).dt.to_period('M')

    # Calculate prevalence by period
    prevalence = samples_df.groupby(['period', gene_col]).size().unstack(fill_value=0)

    # Normalize to percentage
    total_per_period = samples_df.groupby('period').size()
    prevalence_pct = prevalence.div(total_per_period, axis=0) * 100

    return prevalence_pct


def detect_emerging_resistance(historical_df, new_samples_df):
    '''Flag novel or increasing resistance patterns

    Alerts for:
    1. New AMR gene not seen before
    2. Significant increase in prevalence
    3. New combinations of resistance
    '''
    historical_genes = set(historical_df['Gene symbol'].unique())
    new_genes = set(new_samples_df['Gene symbol'].unique())

    novel = new_genes - historical_genes

    if novel:
        print(f'ALERT: Novel resistance genes detected: {novel}')

    return novel

# Drug-gene relationships for interpretation
AMR_INTERPRETATION = {
    'bla_CTX-M': {
        'class': 'Beta-lactam',
        'affects': ['Cephalosporins (3rd gen)', 'Penicillins'],
        'clinical': 'ESBL producer - avoid cephalosporins'
    },
    'bla_KPC': {
        'class': 'Beta-lactam',
        'affects': ['Carbapenems', 'Cephalosporins', 'Penicillins'],
        'clinical': 'Carbapenemase - limited treatment options'
    },
    'mcr-1': {
        'class': 'Polymyxin',
        'affects': ['Colistin'],
        'clinical': 'Plasmid-mediated colistin resistance - critical'
    },
    'vanA': {
        'class': 'Glycopeptide',
        'affects': ['Vancomycin', 'Teicoplanin'],
        'clinical': 'VRE - infection control measures required'
    }
}

def interpret_amr_profile(genes):
    '''Generate clinical interpretation of AMR profile'''
    interpretations = []

    for gene in genes:
        for pattern, info in AMR_INTERPRETATION.items():
            if pattern in gene:
                interpretations.append({
                    'gene': gene,
                    **info
                })
                break

    return interpretations

def generate_surveillance_report(samples_df, period='month'):
    '''Generate AMR surveillance summary report

    Standard surveillance metrics:
    - Prevalence by drug class
    - Trends over time
    - Geographic distribution
    - Emerging threats
    '''
    report = {
        'period': period,
        'total_samples': len(samples_df['sample_id'].unique()),
        'total_amr_genes': samples_df['Gene symbol'].nunique()
    }

    # Prevalence by class
    class_counts = samples_df.groupby('Class')['sample_id'].nunique()
    report['prevalence_by_class'] = (class_counts / report['total_samples'] * 100).to_dict()

    # Critical resistance
    critical = ['Carbapenem', 'Colistin', 'Vancomycin']
    for drug in critical:
        matching = samples_df[samples_df['Class'].str.contains(drug, case=False, na=False)]
        report[f'{drug.lower()}_resistance'] = len(matching['sample_id'].unique())

    return report