Inside a MAFI Assessment

What follows is a walkthrough of a MAFI assessment, using Ganetespib - an HSP90 inhibitor developed for advanced non-small cell lung cancer - as the demonstration case.

MAFI's assessment found a programme whose mechanism is structurally fragile across multiple layers. Ganetespib carries no Robust dependencies, concentrates its critical failure points in the biological layer, and positions more closely to failed comparable programmes than successful ones.

The walkthrough draws on two report types. The MDDR assesses a programme's mechanism in absolute terms. The MBR positions that profile against comparable programmes.

The six panels that follow move from the programme's dependency architecture, through its fragility assessment and critical failure points, to its comparative position against reference populations. Each panel shows an actual output from the Ganetespib reports, with a short interpretation of what it found.

Mechanistic Due Dilignce Report

  • Dependency Architecture

    Mechanistic Due Diligence Report

    Ganetespib's dependencies are disproportionately concentrated in Core classifications, with zero Peripheral Biological dependencies. Every biological assumption the programme makes is treated as core to the mechanism.

    Finding: Core concentration at 63.6%, zero Peripheral Biological. Implication: Every biological assumption is core.

  • Fragility Assessment

    Mechanistic Due Diligence Report

    Ganetespib has no Robust dependencies. Four of its eleven dependencies resolve as Brittle; the remaining seven as Conditional. The mechanism carries no assumption that meets the criteria for well-supported, context-independent resilience.

    Finding: No Robust dependencies; four Brittle, seven Conditional.
    Implication: No assumption meets the Robust criteria.

  • Critical Failure Nodes

    Mechanistic Due Diligence Report

    Ganetespib's four Critical Failure Nodes are concentrated in the Biological layer and form a sequential chain — from tumour-cell dependence on HSP90 client proteins through to the persistence of the antitumour effect in the presence of resistance mechanisms. The most consequential points of failure lie in the biology itself.

    Finding: Four Critical Failure Nodes, Biological layer, sequential chain.
    Implication: The most consequential failure points are biological.

Mechanistic Benchmarking Report

  • Dependency Layer Distribution by Comparison Group

    Mechanistic Benchmarking Report

    Ganetespib's dependency architecture is unusually weighted toward Translational assumptions - 36.4% of its dependencies sit in the Translational layer, the highest proportion across all six reference populations. Its Biological proportion is correspondingly depressed at 36.4%, the lowest across all comparison groups. The programme carries far more of its mechanistic burden in translation than its peers.

    Finding: 36.4% Translational (highest across all reference populations); 36.4% Biological (lowest across all comparison groups).

    Implication: The programme carries far more of its mechanistic burden in translation than its peers.

  • Historical Failure Prevalence

    Mechanistic Benchmarking Report

    45.5% of Ganetespib's dependencies fall within the High Historical Failure Prevalence tier - more than twice the comparable cohort's 17.4% - and none fall within the Low tier. The programme's dependency classes are disproportionately drawn from those with an established history of failure.

    Finding: 45.5% High; 0% Low.

    Implication: Dependency classes drawn from historically failure-prone categories.

  • Integrated Fragility

    Mechanistic Benchmarking Report

    Ganetespib has no Robust dependencies, compared with approximately 10% among successful comparable programmes. Its 36.4% Brittle proportion aligns more closely with the failed comparable cohort (37.9%) than the successful comparable cohort (21.9%). The programme's overall fragility profile resembles the profiles associated with failure more closely than those associated with success.

    Finding: No Robust; 36.4% Brittle.

    Implication: Profile resembles failed comparables more than successful ones.