Review Papers Claim by Claim

• Strength: Untargeted metabolomics can produce an unbiased metabolic fingerprint that points to specific enzymes or pathways (examples: eflornithine → ODC; fosmidomycin → DXR/IspD) Creek & Barrett 2013
• Limitation: Metabolite identification (isomers, missing standards) and non-specific stress signatures remain the major bottlenecks and risk false leads Creek & Barrett 2013 - ID challenges
• Best practice recommendation: multifactorial experimental design — dose-response, timecourse, host-parasite separation, orthogonal assays (targeted metabolomics, fluxomics, genetics/proteomics) to separate primary from secondary effects Creek & Barrett 2013 - recommendations

• Eflornithine (T. brucei) — untargeted HILIC-Orbitrap LC-MS: accumulation of ornithine, depletion of putrescine/spermidine → confirmed ODC inhibition Eflornithine → ODC (reviewed)
• Atovaquone / CK-2-68 (P. falciparum) — targeted LC-MS: accumulation of dihydroorotate and carbamoyl-aspartate → links to mitochondrial bc1 / DHODH inhibition cascade Atovaquone metabolic signature
• Fosmidomycin (P. falciparum) — targeted profiling of non-mevalonate isoprenoid intermediates → DXR and IspD implicated (two-step consequence discovered) Fosmidomycin targets (reviewed)

1. Sampling & quenching: Review correctly flags need for rapid quenching (cold) and careful parasite isolation (e.g., saponin lysis, MACS for P. falciparum) to avoid host metabolite contamination — these steps are essential but under-standardized across labs Sampling & quenching caution
2. Analytical platform choices: HILIC + high-resolution Orbitrap/TOF maximizes polar metabolite coverage; reversed-phase for lipids — correct choice but requires multi-platform workflows for near-complete coverage; instrumentation/column choices must be reported in detail for reproducibility Platform guidance
3. Data processing & identification: The review lists XCMS, MZmine, mzMatch, IDEOM and stresses the identification bottleneck — this remains the key barrier (MS/MS libraries, retention-time matched standards, spectral repositories needed) Data tools & ID bottleneck

• Non-metabolic MOAs: Drugs acting via protein–protein interactions, membrane disruption, ion channels or DNA intercalation may not produce a clear, interpretable metabolite fingerprint; negative metabolomics result is not proof of absence of specific MOA Metabolomics limits for non-metabolic MOAs
• Culture-media confounding: High folate in T. brucei medium (HMI11) can mask antifolate activity; thus media design is a hidden variable that can produce false negatives/altered pharmacology Media confounding example
• Host–parasite metabolic interplay: For intracellular parasites, host metabolic responses to drug or infection may dominate signals if parasites are not separated cleanly; host-only controls and parasite purification are essential Host-parasite confounding

1. Standardize and publish sampling/quenching SOPs for each parasite model (including media composition) and require full metadata with metabolomics datasets.
2. Adopt multi-platform metabolomics (HILIC + RP + targeted MS/MS) for discovery → confirmation pipelines and deposit MS/MS spectra in public repositories.
3. Integrate metabolomics with targeted flux analysis (13C tracing) and genetics (knockdown/overexpression) to test causality rather than correlation.
4. Use computational pathway priors and probabilistic formula assignment (retention-time prediction, network-aware ID) to raise confidence in putative IDs before investing in standards.

I judge the central claims of the review (that metabolomics can identify many metabolic MOAs and that identification + orthogonal validation are required) as supported by multiple primary studies cited in the review. Confidence: moderate-high for metabolic-enzyme targeting drugs; low for drugs whose primary MOA is non-metabolic. Disproof/falsification would require systematic demonstration that distinct MOAs produce indistinguishable metabolomic signatures even under idealized, high-quality experiments — such a falsification has not been shown in the literature summarized by the review Confidence & falsifiability (review)