-
Framing and scope mismatch: the manuscript advertises a lab-strain vs field-isolate comparative study, but the Results present only field-isolate trajectory/peak findings because lab-strain cells are insufficient for robust trajectory/peak analyses (Sec. 1; Sec. 2.5; Sec. 3.1; Sec. 4). This is currently misleading and obscures the true contribution (field-only in vivo analysis).
Recommendation: Make a clear scope decision and align the entire manuscript. (A) If a lab-vs-field comparison is feasible at any level, add a dedicated Results subsection (Sec. 3) that reports what can be supported (e.g., cell/stage distributions, expression of a small set of key regulators, reference-mapping to an atlas), and ensure Sec. 2.5 describes only analyses actually performed. (B) If not feasible, refocus the Abstract/Title/Introduction (Sec. 1) and Conclusions (Sec. 4) to present this as a field-isolate study; substantially shorten/remove Sec. 2.5; and explicitly state as a limitation that lab-strain trajectories/peaks could not be inferred due to insufficient stage coverage/cell counts, with the quantitative breakdown reported in Sec. 3.1.
-
Pseudotime methodology is internally inconsistent and under-specified, undermining all downstream timing/window/peak claims. Methods describe Monocle 3 pseudotime (Sec. 2.3.3), while Results describe Diffusion Pseudotime (DPT) (Sec. 3.1) and figures/captions reference DPT. Key parameters (normalization, PCs used, neighbor graph construction, diffusion components, root choice, branch handling) are not reported in a reproducible way (Sec. 2.3.1–2.3.4; Sec. 3.1).
Recommendation: Unify the pipeline description and implementation. In Sec. 2.3, state unambiguously which algorithm produced the pseudotime used for all reported peaks and windows (DPT, Monocle 3, or both with one chosen for main results). Provide essential parameters: software package + version, input layer (raw/log-normalized/scaled), number of HVGs and PCs, kNN $k$, diffusion components (if DPT), UMAP settings, root-selection rule (and justification), and how branching was handled (single lineage vs principal graph vs branch-specific pseudotime). Update Sec. 3.1 to match. If both methods were tried, report concordance (e.g., rank correlation; stability of peak locations) and move comparisons to Supplement.
-
Trajectory plausibility and rooting are not convincingly established given sparse asexual anchors and stage imbalance. The dataset is described as gametocyte-dominated with relatively few late rings/early trophozoites anchoring the start of the trajectory (Sec. 3.1). With limited intermediate asexual coverage, there is a risk that the inferred “asexual $\rightarrow$ sexual” continuity is driven by patient/batch structure or embedding artifacts rather than biology.
Recommendation: Strengthen validation and robustness checks in Sec. 3.1. (i) Report counts per annotated stage (including gametocyte I–V; male/female if available) overall and per patient (table + simple stacked bars). (ii) Show continuity diagnostics: neighbor-graph connectivity between annotated stages; whether asexual cells connect to gametocyte populations through intermediate states vs a narrow bridge. (iii) Test robustness to (a) sub-sampling the small asexual set, (b) alternative root choices, and (c) re-running pseudotime per patient or after integration/batch correction. If robust asexual$\rightarrow$sexual ordering cannot be supported, narrow the biological claim to within-gametocyte development (I$\rightarrow$V) and avoid implying capture of commitment if the relevant pre-commitment stages are missing.
-
Marker-based trajectory validation is currently contradicted by the manuscript/figures. Sec. 3.1 claims validation using KAHRP and Pfs16 trends, but the Figure 1 caption states KAHRP and Pfs16 expression data were not found. Additional markers mentioned (e.g., MSP1 in Sec. 2.3.3) are not consistently shown/used. This raises concerns about gene-ID mapping, filtering, and the validity of the trajectory interpretation.
Recommendation: Reconcile marker availability and re-validate the trajectory with markers that are demonstrably present in the expression matrix. Concretely: (i) fix gene identifier mapping/aliases (PF3D7\_… format consistency; PlasmoDB version) and ensure plotting uses correct IDs; (ii) update Sec. 3.1 and Fig. 1 to include a panel of established markers across rings/trophozoites/schizonts (if present) and early/late gametocytes (and sex-specific markers if available), with smoothed expression vs pseudotime; (iii) if key markers were filtered out by expression thresholds, revise filtering or explicitly justify it and discuss implications (especially for sparse TFs like AP2-G). Remove or correct any claims that rely on missing marker plots.
-
The definition of the “trophozoite-to-gametocyte transition window” is too broad and inconsistently aligned with the stated goal of peaks ‘immediately preceding’ a transition. In Sec. 3.2 the window is defined as spanning from the median early trophozoite pseudotime ($\sim0.028$) to the median gametocyte pseudotime ($\sim0.794$), which can cover most of the trajectory and risks selecting late gametocyte programs rather than commitment/transition regulators (Sec. 2.3.4; Sec. 2.4.2–2.4.3; Sec. 3.2).
Recommendation: Redefine transitions locally and consistently with Methods. Options: (i) define a narrow band where stage label probabilities/fractions cross (e.g., where gametocyte fraction rises from $10\% \rightarrow 90\%$); (ii) define a boundary between early gametocyte (I/II) and trophozoite-like cells if those labels exist; (iii) if commitment is not captured, reframe the analysis around gametocyte stage transitions (I$\rightarrow$II$\rightarrow\ldots\rightarrow$V) and define corresponding local windows. Report the number of cells and full pseudotime distributions used to compute any medians (violin/histogram). Then re-run or re-filter peak calls with the updated window and interpret candidates accordingly.
-
The transient peak-detection method is central but insufficiently specified and internally inconsistent, limiting reproducibility and credibility (Sec. 2.4 vs Sec. 3.2). Parameters are described as examples/placeholders in Methods (X-fold, Y-SD, smoothing window, transience threshold) while Results appear to apply fixed values; ‘infinite fold-change’ is reported for AP2-G due to baseline$\approx0$, but division-by-zero/near-zero handling is not defined (Sec. 2.4.2; Sec. 3.2.1). The “mean/std of smoothed baseline expression” criterion is also not mathematically well-defined as written if baseline is a single percentile/median scalar.
Recommendation: Rewrite Sec. 2.4 to be fully operational and match Sec. 3.2 exactly. Specify: (i) smoothing approach (rolling mean/LOESS/GAM), window size (in cells or pseudotime units), and boundary handling; (ii) baseline definition (which percentile/median), and how zeros are treated (pseudocount floor; or use additive difference and/or fraction-expressing criteria instead of fold-change when baseline$\approx0$); (iii) exact thresholds used for peak calling (X-fold, Y-SD) with a clear definition of the distribution used to compute mean/std; (iv) exact transience criterion (and whether it is relative to total pseudotime vs stage duration—make Methods/Results consistent); (v) how multiple peaks per gene are handled. Provide code/pseudocode and, ideally, a sensitivity analysis (Supplement is fine) showing stability of top candidates under reasonable parameter perturbations.
-
Patient-specific heterogeneity claims are not statistically supported and may be confounded by unequal sampling, stage composition differences, and patient/batch effects. The current approach (Fig. 3 stacked proportions within peak windows) can reflect who has cells in that pseudotime region rather than true differential activation (Sec. 3.3–3.4; Sec. 4.2). Uncertainty estimates and formal enrichment tests are missing, and underlying counts are not shown.
Recommendation: Upgrade Sec. 3.1 and Sec. 3.3 to quantify and control confounding. (i) Report per-patient cell counts overall and by stage/pseudotime segment. (ii) For each peak window, report raw counts (total and per patient) and add uncertainty (bootstrap CIs or Bayesian intervals for proportions). (iii) Test enrichment formally against an appropriate null: e.g., permutation of patient labels within local pseudotime neighborhoods; Fisher’s exact/chi-squared comparing peak-window vs matched-background window; or logistic/multinomial regression where ‘in peak window’ is the outcome and patient + stage (or local pseudotime density) are covariates. (iv) Where possible, show within-stage comparisons (e.g., among early gametocytes only: fraction AP2-G$^+$ by patient) to reduce compositional bias. Calibrate language in Sec. 4.2 to distinguish statistically supported heterogeneity from descriptive patterns given $n=4$ patients.
-
Biological interpretation of PP2C and FIKK as transient regulators of commitment/maturation is currently under-supported given the broad transition window and limited demonstration that these profiles are truly transient (peaked) rather than simply stage-enriched/monotonic with pseudotime (Sec. 3.2.2; Sec. 4.2).
Recommendation: Strengthen evidence and contextualization. In Sec. 3.2.2, show expression-vs-pseudotime curves for AP2-G, PP2C, and FIKK with the called peak window highlighted, and quantify ‘transience’ (e.g., peak width; peak-to-baseline difference; fraction of expressing cells). Cross-reference external evidence: prior bulk/scRNA datasets or PlasmoDB stage-enrichment annotations for these genes. If the analysis is more consistent with gametocyte stage progression than commitment, adjust the claim accordingly and discuss alternative interpretations.
-
Core preprocessing/QC and integration details are insufficient for reproducibility and for interpreting whether technical effects drive pseudotime/peaks (Sec. 2.1–2.3). It is unclear what cell filtering thresholds were used (genes/UMIs, mitochondrial content), whether ambient RNA/doublets were addressed, and whether batch/patient integration or correction was performed.
Recommendation: Expand Sec. 2.1–2.3 with exact QC and normalization details: cell and gene filters (with thresholds), normalization/scaling (and covariates regressed), HVG selection parameters, and whether/how patient/batch effects were corrected (method + parameters). Provide summary QC metrics in Results/Supplement (UMIs/genes per cell distributions; post-filter counts per sample). If the workflow follows Dogga et al. (2024) or another source, state precisely what was reproduced and what was modified.