Bio Clinical Databases Msi Detection

Version Compatibility

Reference examples tested with: MSIsensor-pro 1.2+, MSIsensor 0.6+, MANTIS 1.0.5+, samtools 1.19+, mSINGS 5.6+, pandas 2.2+, cyvcf2 0.30+. FDA pembrolizumab MSI-H / dMMR pan-tumor approval is from 2017 (Le 2015 NEJM; KEYNOTE-016/164/158); approval extended to colorectal first-line in 2020.

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

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

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying. MSIsensor-pro replaces MSIsensor for tumor-only assays; MSIsensor-ct is the bTMB-equivalent for ctDNA panels.

MSI Detection; The Companion ICI Biomarker to TMB

'Detect MSI status from this somatic sequencing data' -> Profile microsatellite instability across canonical loci (Bethesda 5 panel + extended NGS-derived sites); classify MSI-H / MSS / MSI-L per Bethesda / FDA / KEYNOTE convention.

• CLI (recommended tumor-only): msisensor-pro msi -d microsatellites.list -t tumor.bam -o msi_out -b 16
• CLI (paired tumor-normal): msisensor msi -d microsatellites.list -n normal.bam -t tumor.bam -o msi_out
• CLI (ctDNA / blood MSI): msisensor-ct ...
• CLI (older WES standard): mantis -t tumor.bam -n normal.bam -b targets.bed --threads 8

The Regulatory and Trial Landscape

MSI vs dMMR vs TMB-H: The Conceptual Hierarchy

MSI-H + TMB-H overlap (Chalmers 2017 Genome Med 9:34):

• ~83% of MSI-H tumors are TMB-H.
• ~16% of TMB-H solid tumors are MSI-H.
• Sha 2020 Cancer Discov: MSI-H is the more established dMMR biomarker for ICI decisions; TMB-H not additive.

POLE-exo vs MMR-D:

• POLE-exo (SBS10a/10b): hypermutator (100-300 mut/Mb pure); typically MSI-stable.
• MMR-D (SBS6/15/26/44 + ID1/2): 30-50 mut/Mb typical; MSI-H.
• POLE-exo + MMR-D (SBS14 + SBS20): ultra-hypermutator >=500 mut/Mb; MSI-H.

Tool Taxonomy

Operational consensus 2024-2026:

• Tumor + paired normal WES: MSIsensor or MANTIS.
• Tumor-only assay (commercial panels, often unpaired): MSIsensor-pro with reference baseline.
• ctDNA / liquid biopsy: MSIsensor-ct.
• Lynch screening: IHC FIRST (rules out 90%+); MSI-PCR / NGS confirmatory.

Decision Tree by Scenario

Bethesda Panel and Modern NGS-Derived Loci

The original NCI/Bethesda reference panel (Boland 1998) used BAT-25 and BAT-26 plus three dinucleotide markers (D2S123, D5S346, D17S250); >= 2 of 5 loci unstable -> MSI-H. Modern PCR assays use the mononucleotide pentaplex (the current clinical standard), which replaced the dinucleotide markers for improved cross-population specificity:

• BAT-25 (chr4)
• BAT-26 (chr2)
• NR-21 (chr14)
• NR-24 (chr2)
• MONO-27 (chr2)

NGS-based MSI panels use 50-1000+ microsatellite loci. MSI-H requires unstable status at >=40% of tested loci typically (varies by panel calibration).

Standard Workflow: MSIsensor-pro Tumor-Only

Goal: Compute MSI status from tumor-only WES/panel.

Approach: Generate baseline from population reference; compare patient tumor.

# Generate microsatellite list from reference genome (one-time)
msisensor-pro scan -d /reference/GRCh38.fa -o microsatellites.list -p 1 -m 5

# Generate baseline from N normal control samples (one-time per panel)
msisensor-pro baseline -d microsatellites.list -i normal_samples.list -o baseline.list -b 16

# Score tumor sample. The `-i sample_id` flag is uncommon: in typical msisensor-pro
# usage the sample identifier is derived from the BAM file -- verify the flag set
# against `msisensor-pro pro --help` for the installed release.
msisensor-pro pro \
    -d microsatellites.list \
    -t tumor.bam \
    -o msi_output \
    -b 16 \
    --baseline baseline.list

# Output: msi_output_all (raw); msi_output_unstable (unstable loci); msi_output.txt (summary)
# Critical column: %_unstable. Threshold MSI-H typically >= 20-30% depending on panel.

Paired Tumor-Normal MSIsensor

msisensor msi \
    -d microsatellites.list \
    -n normal.bam \
    -t tumor.bam \
    -o msi_paired_out \
    -b 16

# Output: %_unstable in paired comparison
# MSI-H threshold: >= 20% by FoCR guidance; varies 10-30% across studies

MANTIS Step-wise Difference

mantis.py \
    -t tumor.bam \
    -n normal.bam \
    -b microsatellite_targets.bed \
    --threads 8 \
    -o mantis_output

# Output: mantis_output.kmer_counts (raw), mantis_output (status)
# Threshold MSI-H: stepwise difference > 0.4 (default)

MSI-H Classification Logic

import pandas as pd


def classify_msi(unstable_percentage, panel_calibrated_cutoff=20.0):
    '''Classify MSI status from percentage of unstable loci.

    Bethesda PCR: >=2 of 5 unstable -> MSI-H (40% loci)
    NGS: panel-specific cutoffs typically 10-30%
    Concordance: MSI-PCR + IHC + NGS should agree (FoCR)
    '''
    if unstable_percentage >= panel_calibrated_cutoff:
        return 'MSI-H'
    elif unstable_percentage >= panel_calibrated_cutoff / 2:
        return 'MSI-L (intermediate; treat as MSS clinically per FDA)'
    else:
        return 'MSS'


def msi_lynch_workflow(msi_status, ihc_results, mlh1_methylation_status, germline_test):
    '''Standard Lynch syndrome workflow.

    Args:
        msi_status: 'MSI-H' / 'MSS' / 'MSI-L'
        ihc_results: dict {MLH1: 'retained' or 'loss', MSH2, MSH6, PMS2}
        mlh1_methylation_status: 'methylated' (sporadic) / 'unmethylated' (Lynch suspect)
        germline_test: 'positive' / 'negative' / 'not_performed'
    '''
    if msi_status != 'MSI-H':
        return 'No further Lynch screening indicated'

    ihc_loss = [gene for gene, status in ihc_results.items() if status == 'loss']
    if not ihc_loss:
        return 'MSI-H with retained IHC; consider Lynch with germline testing'

    if 'MLH1' in ihc_loss:
        if mlh1_methylation_status == 'methylated':
            return 'Sporadic MSI-H (MLH1 hypermethylation); not Lynch'
        elif mlh1_methylation_status == 'unmethylated':
            return 'Lynch suspect (MLH1 loss without methylation); proceed with germline testing'
        else:
            return 'MLH1 loss; perform methylation test'

    return f'MSH2/6/PMS2 loss ({", ".join(ihc_loss)}); strong Lynch suspect; germline testing'


def msi_tmb_ici_decision(msi_status, tmb_value, tumor_type=None, dmmr_ihc=None):
    '''Integrated ICI eligibility from MSI + TMB.

    Sha 2020: MSI-H is primary biomarker; TMB-H not additive.
    McGrail 2021: TMB-H NOT endorsed for breast/prostate/glioma alone.
    '''
    msi_high = msi_status == 'MSI-H'
    dmmr_positive = dmmr_ihc == 'positive'
    tmb_h = tmb_value >= 10

    if msi_high or dmmr_positive:
        return ('ICI eligible: MSI-H or dMMR (FDA pembrolizumab 2017 pan-tumor; KEYNOTE-016/164/158); '
                'TMB-H is not additive (Sha 2020).')
    if tmb_h and tumor_type and tumor_type.lower() in ('breast', 'prostate', 'glioma'):
        return ('TMB-H but tumor type excluded by ESMO 2024 / McGrail 2021. '
                'Consider tumor-type-specific cutoff.')
    if tmb_h:
        return 'TMB-H pan-tumor (FDA pembrolizumab 2020); ICI eligible.'
    return 'MSS + TMB-low. Standard chemo per tumor type.'

Per-Operation Failure Modes

1. Tumor-only with paired-normal tool

• Trigger: Run MSIsensor on tumor-only BAM.
• Mechanism: MSIsensor requires paired normal for baseline comparison.
• Symptom: Tool errors or produces unstable noisy result.
• Fix: Use MSIsensor-pro for tumor-only; or use mSINGS background-panel approach.

2. Panel size too small

• Trigger: 30-locus panel called MSI-H based on 20% threshold (= 6 unstable loci).
• Mechanism: Small panel + stochastic unstable rates produce high false-positive rates.
• Symptom: False-positive MSI-H in WES-comparable panels with < 50 microsatellite loci.
• Fix: Validate panel calibration with reference cohort; use panel-specific cutoff; minimum 50 informative loci.

3. IHC vs MSI discordance not investigated

• Trigger: IHC retains all four MMR proteins; MSI-H by sequencing.
• Mechanism: IHC may miss subtle loss; MSI may include MSH6-only subtype (more variable); rare germline POLE+MMR ultra-hypermutators show MSI.
• Symptom: Apparent discordance; classification ambiguous.
• Fix: Cross-check with germline MMR sequencing; check for POLE-exo on Sigprofiler.

4. MSI-H + Lynch syndrome confusion

• Trigger: Report MSI-H tumor as "Lynch syndrome".
• Mechanism: ~50% of MSI-H CRC is sporadic (MLH1 hypermethylation, not germline Lynch).
• Symptom: Incorrect family counseling; wrong screening.
• Fix: Apply IHC + MLH1 methylation + germline testing workflow.

5. POLE-exo hypermutator labeled MSI

• Trigger: Tumor with 200 mut/Mb POLE-exo signature labeled MSI-H.
• Mechanism: Pure POLE-exo causes hypermutator WITHOUT MSI (different repair mechanism); apparent MSI-H call may be a false positive in high-mutation context.
• Symptom: Misclassification; ICI eligibility still positive but for different mechanism.
• Fix: Run Sigprofiler signatures (SBS10a/10b vs SBS6/15/26/44); confirm POLE-exo via SBS10 contribution.

6. ctDNA MSI without sufficient tumor fraction

• Trigger: Run MSIsensor-ct on cfDNA with <1% tumor fraction.
• Mechanism: Low ctDNA fraction produces noise-dominated unstable locus counts.
• Symptom: False-negative or unstable MSI call.
• Fix: Estimate tumor fraction first (ichorCNA); require >= 3% for reliable cfDNA MSI.

7. Universal screening missed

• Trigger: CRC patient < 70 yr without IHC / MSI.
• Mechanism: NCCN / ACG universal Lynch screening required; without it, Lynch syndrome undiagnosed.
• Symptom: Family loses screening benefit.
• Fix: Universal IHC + MSI on all CRC < 70; institute reflex testing.

8. MSI-L treated as actionable

• Trigger: Report MSI-L (intermediate) as ICI-eligible.
• Mechanism: FDA approval specifies MSI-H; MSI-L = MSS clinically.
• Symptom: ICI given on insufficient indication; reimbursement issues.
• Fix: Apply MSI-H threshold strictly per FDA; MSI-L = MSS.