[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EEFQDYH7ER7B3S27RYCJD2":3,"skill-same:01M7EEFQDYH7ER7B3S27RYCJD2":62},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":30,"commitSha":31,"skillDirectory":32,"forksCount":14,"repositoryHtmlUrl":33,"repositoryHomepageUrl":33,"ownerHtmlUrl":34,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":35,"install":36,"resources":49},"01M7EEFQDYH7ER7B3S27RYCJD2","bio-microbiome-differential-abundance",null,"Tests which individual taxa differ between groups on an amplicon ASV\u002Ffeature table (phyloseq) using compositionally-aware methods - ALDEx2 (Dirichlet-MC CLR, conservative), ANCOM-BC2\u002FANCOMBC (sampling-fraction bias correction, structural zeros, passed_ss, default p_adj_method=holm), MaAsLin2\u002FMaAsLin3 (multivariable GLM, random effects, prevalence\u002Fabundance split), LinDA (CLR mixed-model regression), ZicoSeq (permutation FDR), LEfSe, and q2-composition ancombc. Covers why the hit list depends more on the DA tool than the biology (Nearing benchmark) so the deliverable is a CONSENSUS of >=2 tools, why a relative change is not absolute without a load anchor, the prevalence-filter knob, BH\u002FFDR plus an effect-size floor, and why DESeq2\u002FedgeR misfire here. Use when finding differentially abundant taxa, handling covariates or longitudinal designs, or choosing a method. Whole-community diversity -> diversity-analysis; shotgun DA -> metagenomics\u002Fmetagenome-visualization; CoDA theory -> metagenomics\u002Fabundance-estimation","Mr-Nobody-Anonymous\u002FAll-skills","Mr-Nobody-Anonymous","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F184253119?v=4","01M7AEXFJHKTYQW5R0ZCQQTWFV","anonymous","01M7C00M3GQR4QKP9GZA42GFAK",0,"MIT","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-microbiome-differential-abundance\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: bio-microbiome-differential-abundance\ndescription: Tests which individual taxa differ between groups on an amplicon ASV\u002Ffeature table (phyloseq) using compositionally-aware methods - ALDEx2 (Dirichlet-MC CLR, conservative), ANCOM-BC2\u002FANCOMBC (sampling-fraction bias correction, structural zeros, passed_ss, default p_adj_method=holm), MaAsLin2\u002FMaAsLin3 (multivariable GLM, random effects, prevalence\u002Fabundance split), LinDA (CLR mixed-model regression), ZicoSeq (permutation FDR), LEfSe, and q2-composition ancombc. Covers why the hit list depends more on the DA tool than the biology (Nearing benchmark) so the deliverable is a CONSENSUS of >=2 tools, why a relative change is not absolute without a load anchor, the prevalence-filter knob, BH\u002FFDR plus an effect-size floor, and why DESeq2\u002FedgeR misfire here. Use when finding differentially abundant taxa, handling covariates or longitudinal designs, or choosing a method. Whole-community diversity -> diversity-analysis; shotgun DA -> metagenomics\u002Fmetagenome-visualization; CoDA theory -> metagenomics\u002Fabundance-estimation\ntool_type: r\nprimary_tool: ALDEx2\nsource: \"https:\u002F\u002Fgithub.com\u002FGPTomics\u002FbioSkills\"\nsource_repository: \"GPTomics\u002FbioSkills\"\nsource_path: \"microbiome\u002Fdifferential-abundance\u002FSKILL.md\"\nlicense: \"MIT\"\nimported_at: \"2026-09-20\"\n---\n\n## Version Compatibility\n\nReference examples tested with: ALDEx2 1.34+, ANCOMBC 2.4+, Maaslin2 1.16+, MicrobiomeStat 1.2+ (LinDA), GUniFrac 1.8+ (ZicoSeq), phyloseq 1.46+.\n\nBefore using code patterns, verify installed versions match. If versions differ:\n- R: `packageVersion('\u003Cpkg>')` then `?function_name` to verify parameters\n\nIf code throws ImportError, AttributeError, or TypeError, introspect the installed\npackage and adapt the example to match the actual API rather than retrying.\n\nANCOM-BC2 changed argument names between `ancombc()` and `ancombc2()`, and its default `p_adj_method` is `holm`, not `BH` - confirm both against the installed version. The MaAsLin3 `maaslin3()` API differs from MaAsLin2's `Maaslin2()`.\n\n# Differential Abundance Testing\n\n**\"Find which taxa differ between my groups\"** -> Run two or more compositionally-aware DA tools and report their consensus - because the significant-taxa list is a property of the tool as much as of the sample, and a relative-abundance change is not an absolute change.\n- R: `ALDEx2::aldex(counts, conds, test='t', effect=TRUE, denom='all')` then a second tool (`ANCOMBC::ancombc2()` or `MicrobiomeStat::linda()`)\n\nScope: per-taxon DA on an amplicon feature table. Whole-community alpha\u002Fbeta\u002FPERMANOVA -> diversity-analysis. Shotgun profiler-table DA -> metagenomics\u002Fmetagenome-visualization. Shared compositional\u002Fclosure\u002FCLR\u002Fzero theory -> metagenomics\u002Fabundance-estimation. Collapse ASVs to genus\u002Fspecies first -> taxonomy-assignment. QIIME2 CLI route -> qiime2-workflow.\n\n## The Single Most Important Modern Insight -- Which Taxa Are \"Significant\" Depends More on the Tool Than on the Biology\n\nRun ALDEx2, ANCOM-BC2, MaAsLin2, and LinDA on the same ASV table and the four significant-taxa lists overlap but disagree (Nearing 2022 *Nat Commun* 13:342, across 38 datasets). So the deliverable is NOT \"the differential taxa\" - it is the CONSENSUS of >=2 compositionally-aware tools, every tool NAMED: the intersection is high-confidence, the union is exploratory, and a single-tool hit is tentative. Picking the tool with the prettiest volcano is p-hacking by software (uncorrected multiplicity hidden in the method menu). Three corollaries:\n\n1. **A relative-abundance increase is not an absolute increase.** Microbiome counts are compositional - the sequencer fixes the total, so one taxon blooming forces every other taxon's proportion down (the blooming-taxon illusion). \"Taxon X increased\" is a statement about its SHARE unless an external load anchor (spike-in \u002F flow cytometry \u002F qPCR, see metagenomics\u002Fabundance-estimation) or MaAsLin3's absolute-abundance mode licenses an absolute claim.\n2. **Uncorrected Wilcoxon\u002Ft-test on raw relative abundances is wrong twice in one line** - closure (reference-frame) AND multiple testing. But a BH-corrected simple test, honestly labelled as relative, can replicate BETTER than a sophisticated model (Pelto 2025): the forbidden thing is the uncorrected, closure-blind form, not simple tests per se.\n3. **There is no settled best tool.** The benchmarks optimize different criteria, so they rank tools differently. Consensus-of-tools is the only stance that survives all of them.\n\n## The Benchmark Landscape (no settled winner)\n\n| Benchmark | Optimized for | Verdict |\n|-----------|---------------|---------|\n| Nearing 2022 *Nat Commun* 13:342 | cross-method consistency | ALDEx2 + ANCOM-II most consistent and most conservative; LEfSe\u002FedgeR flag far more, agree less |\n| Yang & Chen 2022 *Microbiome* 10:130 | FDR-power balance | ZicoSeq \u002F LinDA \u002F ANCOM-BC-family best |\n| Yang & Chen 2023 *Brief Bioinform* 24:bbac607 | correlated (repeated-measures) designs | use a mixed-model-capable tool (LinDA, MaAsLin2, ANCOM-BC2) |\n| Pelto 2025 *Brief Bioinform* 26(2):bbaf130 | cross-study replicability | elementary BH-corrected methods most replicable; ANCOM-BC2 worst |\n\nReport the disagreement AS the result; verify current best practice against the latest tool docs rather than hard-coding one method.\n\n## Tool Taxonomy\n\n| Tool | Citation | Mechanism \u002F role | When |\n|------|----------|------------------|------|\n| ALDEx2 | Fernandes 2014 *Microbiome* 2:15 | Dirichlet Monte-Carlo posterior + CLR; tests each draw; reports expected effect + BH-adjusted p | conservative two-group anchor; small-to-moderate n |\n| ANCOM-BC2 | Lin & Peddada 2024 *Nat Methods* 21:83 | estimates per-sample sampling fraction and bias-corrects; structural zeros; pseudo-count sensitivity (`passed_ss`) | interpretable LFC + CI; covariates; multi-group |\n| MaAsLin2 | Mallick 2021 *PLoS Comput Biol* 17:e1009442 | general (mixed) linear model on transformed abundance | multivariable \u002F longitudinal \u002F metadata-rich |\n| MaAsLin3 | Nickols 2026 *Nat Methods* 23:554 | splits abundance (level when present) from prevalence (present\u002Fabsent); absolute-abundance mode | prevalence-vs-abundance separation; load data available |\n| LinDA | Zhou 2022 *Genome Biol* 23:95 | CLR regression with mode-based bias correction; asymptotic FDR | large cohorts; fast; native mixed model |\n| ZicoSeq | Yang & Chen 2022 *Microbiome* 10:130 | reference-taxa normalization + permutation FDR; winsorization | covariates; non-parametric permutation p; strong FP control |\n| LEfSe | Segata 2011 *Genome Biol* 12:R60 | Kruskal-Wallis + LDA effect size | exploratory biomarker discovery; NOT a formal FDR-controlled test |\n| DESeq2 | Love 2014 *Genome Biol* 15:550 | RNA-seq median-of-ratios size factor | caveat only; geometric-mean reference dies on sparse zero-heavy tables |\n\n## Decision Tree by Scenario\n\n| Scenario | Recommended | Why |\n|----------|-------------|-----|\n| Two groups, want a trustworthy conservative anchor | ALDEx2 | Dirichlet-MC + CLR; most reproducible\u002Fconservative (Nearing); gate on effect size |\n| Need interpretable LFC + CI, structural zeros, multi-group | ANCOM-BC2 | models and corrects per-sample sampling fraction; global\u002Fpairwise\u002FDunnett\u002Ftrend; `passed_ss` |\n| Large cohort, covariates, speed, mixed model | LinDA | CLR regression + bias mode; asymptotic FDR; fast; random effects in the formula |\n| Covariates + permutation-grounded non-parametric p | ZicoSeq | reference-taxa frame + permutation FDR |\n| Longitudinal \u002F many covariates \u002F flexible GLM | MaAsLin2 | `fixed_effects` + `random_effects`; normalization\u002Ftransform menu |\n| Prevalence-vs-abundance separation or absolute abundance | MaAsLin3 | logistic prevalence model + abundance model; load-data hook |\n| Inside a QIIME2 CLI pipeline | qiime composition ancombc + tabulate\u002Fda-barplot | native artifact flow (v1 ANCOM-BC; go to R for v2 `passed_ss`\u002Fmulti-group) -> qiime2-workflow |\n| Repeated \u002F paired samples | any tool above WITH a random effect | ignoring subject structure is pseudo-replication |\n| ALWAYS | run >=2 of the above, report the consensus | tool choice drives the hit list more than biology (Nearing 2022) |\n| Shotgun species table, not amplicon | -> metagenomics\u002Fmetagenome-visualization | same CoDA theory; different upstream pipeline |\n| Uncorrected t-test\u002FWilcoxon on TSS proportions | DO NOT | closure biases the test and there is no FDR control |\n\n## Filter Before Testing (a modeling knob, not housekeeping)\n\n**Goal:** Drop rare features before testing so the BH denominator is not crushed and log\u002FCLR transforms are well-behaved.\n\n**Approach:** Keep features present in at least 10-25% of samples (and optionally a mean-abundance floor); declare the threshold and confirm the headline result is not knife-edge-sensitive to it. Every tool exposes this (`prv_cut`, `min_prevalence`, `prev.filter`).\n\n```r\nlibrary(phyloseq)\nps \u003C- readRDS('phyloseq_object.rds')\n# prv_cut 0.10: a feature must appear in >= 10% of samples; raising to 0.25 removes more tests\n# (smaller BH correction, more power on survivors) but discards rare-but-real taxa - a declared choice\nkeep \u003C- filter_taxa(ps, function(x) sum(x > 0) >= 0.10 * nsamples(ps), TRUE)\n```\n\n## ALDEx2: The Conservative Floor of the Consensus\n\n**Goal:** Identify taxa that differ between two groups while propagating the sampling uncertainty of low-count features.\n\n**Approach:** Draw `mc.samples` Monte-Carlo instances from a Dirichlet posterior of the counts (this IS the zero handling - no explicit pseudocount), CLR-transform each instance against the geometric mean of all features (`denom='all'`), run the test on every draw, and report the EXPECTED effect size and BH-adjusted p over the draws.\n\n```r\nlibrary(ALDEx2)\ncounts \u003C- as.matrix(otu_table(ps))            # integer counts, taxa in ROWS\nif (!taxa_are_rows(ps)) counts \u003C- t(counts)\ngroups \u003C- as.character(sample_data(ps)$Group)\n\n# mc.samples 128: standard Monte-Carlo draws; 256+ for publication (more stable expected p)\nres \u003C- aldex(counts, groups, mc.samples = 128, test = 't', effect = TRUE, denom = 'all')\n# we.eBH = Welch expected BH-adjusted p (report this, NOT we.ep); wi.eBH = Wilcoxon equivalent\n# effect = median standardized effect = median(diff.btw \u002F max(diff.win)); the primary decision variable\nhits \u003C- res[res$we.eBH \u003C 0.05 & abs(res$effect) > 1, ]   # q AND effect floor (Gloor: gate on effect, not p alone)\n```\n\nGate on effect size AND q, not p alone: with large n trivially small CLR differences become \"significant,\" and Gloor's own guidance is that `|effect| > 1` is a strong ~2-SD signal. For >2 groups use `aldex.kw()`; for covariates the `aldex.glm()` + model.matrix route works but ALDEx2 is weakest here - prefer ANCOM-BC2\u002FLinDA\u002FMaAsLin2 for serious covariate or random-effect modeling.\n\n## ANCOM-BC2: Bias-Corrected LFC With a Sensitivity Safeguard\n\n**Goal:** Estimate an interpretable bias-corrected log-fold-change per taxon, with covariate adjustment, structural-zero handling, and a flag for hits that are hostage to the pseudo-count.\n\n**Approach:** Model log(observed count) as a function of covariates, estimate each sample's log sampling fraction as an offset and subtract it, then refit across a range of pseudo-counts and record how often each q-value flips (`passed_ss`).\n\n```r\nlibrary(ANCOMBC)\nout \u003C- ancombc2(data = ps, fix_formula = 'Group + Age + Sex',\n                rand_formula = NULL,        # '(1 | SubjectID)' for repeated measures - see Failure Modes\n                p_adj_method = 'BH',        # DEFAULT is 'holm'; set 'BH' deliberately for FDR\n                prv_cut = 0.10, lib_cut = 1000,\n                group = 'Group', struc_zero = TRUE, pseudo_sens = TRUE,\n                global = FALSE, pairwise = FALSE, n_cl = 2)\nres \u003C- out$res\n# a confident hit is BOTH significant AND robust to the pseudo-count. ANCOM-BC2 suffixes the\n# diff_\u002Fpassed_ss_ columns with the literal model-matrix coefficient (variable + factor level,\n# verbatim case, e.g. 'Grouptreated') - match it by pattern rather than hard-coding the case.\ndcol \u003C- grep('^diff_Group', names(res), value = TRUE)[1]\nrobust \u003C- res[res[[dcol]] & res[[sub('^diff_', 'passed_ss_', dcol)]], ]\n```\n\n`passed_ss` is the most valuable ANCOM-BC2-specific feature: a CLR\u002Flog model on sparse data is hostage to the zero-replacement constant, and `passed_ss` quantifies that per taxon. A hit with `passed_ss == FALSE` depends on the arbitrary pseudo-count - do not report it as confident. For >2 groups set `global=TRUE` (omnibus), `pairwise=TRUE` (mdFDR-controlled pairs), `dunnet=TRUE`, or `trend=TRUE`; results land in `out$res_global`\u002F`res_pair`\u002F`res_dunn`\u002F`res_trend`.\n\n## LinDA: Fast CLR Regression With Native Mixed Models\n\n**Goal:** Get FDR-controlled log2-fold-changes on a large cohort, including repeated-measures designs, without Monte-Carlo or EM cost.\n\n**Approach:** Fit ordinary linear regression on the CLR-transformed table covariate by covariate, estimate the compositional bias as the mode of the per-feature coefficients and subtract it; a random effect in the formula makes it a linear mixed model.\n\n```r\nlibrary(MicrobiomeStat)\notu \u003C- as.data.frame(otu_table(ps)); if (!taxa_are_rows(ps)) otu \u003C- t(otu)\nmeta \u003C- as.data.frame(sample_data(ps))\nfit \u003C- linda(feature.dat = otu, meta.dat = meta,\n             formula = '~ Group + Age + (1 | SubjectID)',   # random effect -> mixed model\n             feature.dat.type = 'count', prev.filter = 0.10, alpha = 0.05)\nfit$output[[1]]   # names(fit$output) are the model-matrix coefficient columns (e.g. 'Grouptreated' - the factor level keeps its case); per-feature: log2FoldChange, lfcSE, stat, pvalue, padj, reject\n```\n\nLinDA is the natural fast modern entry in a consensus panel and the cleanest route to mixed models. Yang & Chen rate it among the best FDR-power trade-offs.\n\n## MaAsLin2 \u002F MaAsLin3 and ZicoSeq (the rest of the panel)\n\n**Goal:** Fit covariate-rich or longitudinal differential-abundance models, or add a permutation-based panel member, when ALDEx2\u002FANCOM-BC2\u002FLinDA do not cover the design.\n\n**Approach:** Use MaAsLin2\u002F3 for multivariable GLMs with random effects, or ZicoSeq for a non-parametric permutation-FDR test against empirically selected reference taxa.\n\nMaAsLin2 fits a flexible per-feature GLM; its package DEFAULT is TSS + LOG + LM (not CLR), and `random_effects` is the canonical route for longitudinal designs. NOTE the orientation gotcha: it expects features in COLUMNS, samples in rows.\n\n```r\nlibrary(Maaslin2)\nfit \u003C- Maaslin2(input_data = as.data.frame(t(otu)), input_metadata = meta,\n                output = 'maaslin2_out', fixed_effects = c('Group', 'Age'),\n                random_effects = c('SubjectID'),\n                normalization = 'TSS', transform = 'LOG', analysis_method = 'LM',\n                min_prevalence = 0.10, max_significance = 0.05)\n# writes all_results.tsv \u002F significant_results.tsv with columns feature, metadata, coef, pval, qval\n```\n\nMaAsLin3 (`maaslin3()`) splits each feature into an abundance model (level when present) and a logistic prevalence model (present\u002Fabsent) tested jointly, and can ingest total-load measurements for absolute-abundance inference. ZicoSeq (`GUniFrac::ZicoSeq()`) winsorizes, posterior-samples, normalizes against empirically selected reference taxa, and returns permutation FDR (`zc$p.adj.fdr`) - a non-parametric panel member that accepts covariates via `adj.name`.\n\n## Consensus: Intersect the Tools\n\n**Goal:** Convert two or more per-tool hit sets into a confidence-graded result instead of one tool's answer.\n\n**Approach:** Collect the significant feature SETS (BH within each tool), then report the intersection as high-confidence, the union as exploratory, and tabulate, per taxon, how many of N tools agree and which ones. Never pool p-values across tools.\n\n```r\nsig_aldex \u003C- rownames(res)[res$we.eBH \u003C 0.05 & abs(res$effect) > 1]\nsig_linda \u003C- rownames(fit$output[[1]])[fit$output[[1]]$reject]   # [[1]] = the group coefficient (named 'Grouptreated')\nconfident  \u003C- intersect(sig_aldex, sig_linda)   # high-confidence\nexploratory \u003C- union(sig_aldex, sig_linda)       # report with the tool that found each\n```\n\n## Per-Method Failure Modes\n\n### Cherry-picking the tool with the prettiest result\n**Trigger:** running several tools and reporting only the one(s) that flag the favored taxon. **Mechanism:** that is uncorrected multiplicity hidden in the method menu (p-hacking by software). **Symptom:** \"the recommended method found X\" with no mention of the tools that disagreed. **Fix:** decide the panel a priori, report ALL tools, intersect for confident hits, disclose disagreement.\n\n### Uncorrected Wilcoxon\u002Ft-test on relative abundances\n**Trigger:** a per-taxon Wilcoxon\u002Ft-test on TSS proportions with no FDR correction. **Mechanism:** closure makes a naive test call every taxon \"decreased\" when one blooms, and hundreds of uncorrected tests inflate false positives. **Symptom:** dozens of \"significant\" taxa, all in the same direction, no q-values. **Fix:** use a CoDA\u002Freference-frame tool; if a simple test is used, BH-correct it and label the comparison as relative (Pelto 2025).\n\n### Pseudo-replication of repeated measures\n**Trigger:** longitudinal\u002Fpaired samples treated as independent rows. **Mechanism:** fewer independent units than rows inflates significance. **Symptom:** implausibly small p-values on a small subject count. **Fix:** a random effect - ANCOM-BC2 `rand_formula='(1|SubjectID)'`, MaAsLin2 `random_effects='SubjectID'`, LinDA `(1|SubjectID)` in the formula. Cross-check ANCOM-BC2 mixed-model output against LinDA\u002FMaAsLin2 (GitHub issue #111 reported `rand_formula` correctness problems in some versions).\n\n### Prevalence filter set blindly\n**Trigger:** an undeclared prevalence cut, or none at all. **Mechanism:** the cut decides which taxa are even tested and thus the BH landscape - it is a modeling choice. **Symptom:** the hit list changes materially between `prv_cut=0.1` and `0.25`. **Fix:** declare and justify the threshold; confirm the headline result survives moving it.\n\n### Relative change reported as absolute\n**Trigger:** \"taxon X doubled\" from a closed table with no load data. **Mechanism:** one taxon blooming compresses every other proportion. **Symptom:** whole-community \"depletion\" that is really one taxon rising. **Fix:** anchor to load (spike-in\u002Fflow\u002FqPCR) or MaAsLin3 absolute mode; otherwise state the claim is relative.\n\n### DESeq2\u002FedgeR on a sparse 16S table\n**Trigger:** RNA-seq median-of-ratios \u002F TMM on a zero-heavy ASV table. **Mechanism:** the geometric-mean size-factor reference collapses on zeros and the \"most features unchanged\" assumption is violated. **Symptom:** degenerate size factors, errors, or inflated hit counts that disagree with CoDA tools (Nearing). **Fix:** use a compositional tool; if DESeq2 is unavoidable, the `poscounts` estimator is the minimum mitigation - present as a caveat, not a recipe.\n\n### ANCOM-BC2 hit held hostage by the pseudo-count\n**Trigger:** reporting `diff_* == TRUE` without checking `passed_ss_*`. **Mechanism:** significance depends on the arbitrary zero-replacement constant. **Symptom:** a hit that vanishes when the pseudo-count changes. **Fix:** require `diff_* & passed_ss_*` for a confident call.\n\n## Quantitative Thresholds\n\n| Threshold | Source | Rationale |\n|-----------|--------|-----------|\n| Prevalence cut 10-25% (`prv_cut`\u002F`min_prevalence`\u002F`prev.filter`) | Nearing 2022; tool defaults (0.10) | rare features carry little information and crush the BH denominator; declare the value and test sensitivity |\n| BH q \u003C= 0.05 across taxa, within each tool | Benjamini-Hochberg 1995 *JRSS B* 57:289 | hundreds-thousands of features make uncorrected p meaningless; do not pool p across tools |\n| ALDEx2 `|effect| > 1` (with q \u003C= 0.05) | Gloor 2016 *J Comput Graph Stat* 25:971 | effect is a standardized median-ratio; ~2 SD is a strong signal; large n makes trivial diffs \"significant\" |\n| ALDEx2 `mc.samples` = 128 (256+ for publication) | Fernandes 2014 *Microbiome* 2:15 | Monte-Carlo draws; more draws stabilize the expected p |\n| ANCOM-BC2 `passed_ss == TRUE` required | Lin & Peddada 2024 *Nat Methods* 21:83 | flags hits whose significance is hostage to the pseudo-count |\n| Consensus of >=2 compositionally-aware tools | Nearing 2022 *Nat Commun* 13:342 | tool choice drives the hit list more than biology; intersection = confident |\n| ZicoSeq permutations `perm.no` >= 99 | Yang & Chen 2022 *Microbiome* 10:130 | permutation FDR resolution; raise for finer tail p |\n\n## Common Errors\n\n| Error \u002F symptom | Cause | Solution |\n|-----------------|-------|----------|\n| ALDEx2 returns NA effects \u002F errors | proportions or non-integer matrix passed | feed integer COUNTS with taxa in rows |\n| `passed_ss` column missing | `pseudo_sens = FALSE` | set `pseudo_sens = TRUE` (the default) |\n| Far fewer hits than expected | ANCOM-BC2 `p_adj_method` left at `holm` | set `p_adj_method = 'BH'` deliberately if FDR is wanted |\n| MaAsLin2 finds nothing \u002F orientation error | features in rows, not columns | transpose so samples are rows, features columns |\n| Mixed-model hits disagree across tools | `rand_formula` correctness varies by version | cross-check ANCOM-BC2 against LinDA\u002FMaAsLin2 |\n| Tools disagree on the hit list | normal - tool choice drives results | report the consensus and the disagreement, do not cherry-pick |\n| Many \"depleted\" taxa in a host\u002Fplant sample | host mitochondria\u002Fchloroplast 16S inflates the table | filter Mitochondria\u002FChloroplast features (see taxonomy-assignment) before DA |\n| Contaminant ASVs among the hits (low-biomass) | reagent kitome not removed before DA | run decontam upstream with negative controls (amplicon-processing; metagenomics\u002Fcontamination-controls) |\n\n## References\n\n- Fernandes AD, Reid JNS, Macklaim JM, McMurrough TA, Edgell DR, Gloor GB. 2014. Unifying the analysis of high-throughput sequencing datasets: characterizing RNA-seq, 16S rRNA gene sequencing and selective growth experiments by compositional data analysis. *Microbiome* 2:15.\n- Gloor GB, Macklaim JM, Fernandes AD. 2016. Displaying variation in large datasets: plotting a visual summary of effect sizes. *J Comput Graph Stat* 25:971-979.\n- Lin H, Peddada SD. 2020. Analysis of compositions of microbiomes with bias correction (ANCOM-BC). *Nat Commun* 11:3514.\n- Lin H, Peddada SD. 2024. Multigroup analysis of compositions of microbiomes with covariate adjustments and repeated measures (ANCOM-BC2). *Nat Methods* 21:83-91.\n- Mallick H, Rahnavard A, McIver LJ, et al. 2021. Multivariable association discovery in population-scale meta-omics studies. *PLoS Comput Biol* 17:e1009442.\n- Nickols WA, Kuntz T, Shen J, et al. 2026. MaAsLin 3: refining and extending generalized multivariable linear models for meta-omic association discovery. *Nat Methods* 23:554-564.\n- Zhou H, He K, Chen J, Zhang X. 2022. LinDA: linear models for differential abundance analysis of microbiome compositional data. *Genome Biol* 23:95.\n- Yang L, Chen J. 2022. A comprehensive evaluation of microbial differential abundance analysis methods: current status and potential solutions. *Microbiome* 10:130.\n- Yang L, Chen J. 2023. Benchmarking differential abundance analysis methods for correlated microbiome sequencing data. *Brief Bioinform* 24:bbac607.\n- Pelto J, Auranen K, Kujala JV, Lahti L. 2025. Elementary methods provide more replicable results in microbial differential abundance analysis. *Brief Bioinform* 26(2):bbaf130.\n- Nearing JT, Douglas GM, Hayes MG, et al. 2022. Microbiome differential abundance methods produce different results across 38 datasets. *Nat Commun* 13:342.\n- Segata N, Izard J, Waldron L, et al. 2011. Metagenomic biomarker discovery and explanation. *Genome Biol* 12:R60.\n- Love MI, Huber W, Anders S. 2014. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. *Genome Biol* 15:550.\n- Benjamini Y, Hochberg Y. 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. *J R Stat Soc Series B* 57:289-300.\n\n## Related Skills\n\n- diversity-analysis - Whole-community alpha\u002Fbeta\u002FPERMANOVA; answer \"do the communities differ\" before \"which taxa differ\"\n- taxonomy-assignment - Collapse ASVs to genus\u002Fspecies before per-taxon testing\n- amplicon-processing - Produces the ASV feature table tested here\n- qiime2-workflow - The qiime composition ancombc CLI route\n- metagenomics\u002Fabundance-estimation - Shared compositional\u002Fclosure\u002FCLR\u002Fzero\u002Fload-anchor theory\n- metagenomics\u002Fmetagenome-visualization - The same DA mechanics on shotgun profiler tables\n- experimental-design\u002Fmultiple-testing - FDR control and multiplicity across taxa\n","## Version Compatibility\n\nReference examples tested with: ALDEx2 1.34+, ANCOMBC 2.4+, Maaslin2 1.16+, MicrobiomeStat 1.2+ (LinDA), GUniFrac 1.8+ (ZicoSeq), phyloseq 1.46+.\n\nBefore using code patterns, verify installed versions match. If versions differ:\n- R: `packageVersion('\u003Cpkg>')` then `?function_name` to verify parameters\n\nIf code throws ImportError, AttributeError, or TypeError, introspect the installed\npackage and adapt the example to match the actual API rather than retrying.\n\nANCOM-BC2 changed argument names between `ancombc()` and `ancombc2()`, and its default `p_adj_method` is `holm`, not `BH` - confirm both against the installed version. The MaAsLin3 `maaslin3()` API differs from MaAsLin2's `Maaslin2()`.\n\n# Differential Abundance Testing\n\n**\"Find which taxa differ between my groups\"** -> Run two or more compositionally-aware DA tools and report their consensus - because the significant-taxa list is a property of the tool as much as of the sample, and a relative-abundance change is not an absolute change.\n- R: `ALDEx2::aldex(counts, conds, test='t', effect=TRUE, denom='all')` then a second tool (`ANCOMBC::ancombc2()` or `MicrobiomeStat::linda()`)\n\nScope: per-taxon DA on an amplicon feature table. Whole-community alpha\u002Fbeta\u002FPERMANOVA -> diversity-analysis. Shotgun profiler-table DA -> metagenomics\u002Fmetagenome-visualization. Shared compositional\u002Fclosure\u002FCLR\u002Fzero theory -> metagenomics\u002Fabundance-estimation. Collapse ASVs to genus\u002Fspecies first -> taxonomy-assignment. QIIME2 CLI route -> qiime2-workflow.\n\n## The Single Most Important Modern Insight -- Which Taxa Are \"Significant\" Depends More on the Tool Than on the Biology\n\nRun ALDEx2, ANCOM-BC2, MaAsLin2, and LinDA on the same ASV table and the four significant-taxa lists overlap but disagree (Nearing 2022 *Nat Commun* 13:342, across 38 datasets). So the deliverable is NOT \"the differential taxa\" - it is the CONSENSUS of >=2 compositionally-aware tools, every tool NAMED: the intersection is high-confidence, the union is exploratory, and a single-tool hit is tentative. Picking the tool with the prettiest volcano is p-hacking by software (uncorrected multiplicity hidden in the method menu). Three corollaries:\n\n1. **A relative-abundance increase is not an absolute increase.** Microbiome counts are compositional - the sequencer fixes the total, so one taxon blooming forces every other taxon's proportion down (the blooming-taxon illusion). \"Taxon X increased\" is a statement about its SHARE unless an external load anchor (spike-in \u002F flow cytometry \u002F qPCR, see metagenomics\u002Fabundance-estimation) or MaAsLin3's absolute-abundance mode licenses an absolute claim.\n2. **Uncorrected Wilcoxon\u002Ft-test on raw relative abundances is wrong twice in one line** - closure (reference-frame) AND multiple testing. But a BH-corrected simple test, honestly labelled as relative, can replicate BETTER than a sophisticated model (Pelto 2025): the forbidden thing is the uncorrected, closure-blind form, not simple tests per se.\n3. **There is no settled best tool.** The benchmarks optimize different criteria, so they rank tools differently. Consensus-of-tools is the only stance that survives all of them.\n\n## The Benchmark Landscape (no settled winner)\n\n| Benchmark | Optimized for | Verdict |\n|-----------|---------------|---------|\n| Nearing 2022 *Nat Commun* 13:342 | cross-method consistency | ALDEx2 + ANCOM-II most consistent and most conservative; LEfSe\u002FedgeR flag far more, agree less |\n| Yang & Chen 2022 *Microbiome* 10:130 | FDR-power balance | ZicoSeq \u002F LinDA \u002F ANCOM-BC-family best |\n| Yang & Chen 2023 *Brief Bioinform* 24:bbac607 | correlated (repeated-measures) designs | use a mixed-model-capable tool (LinDA, MaAsLin2, ANCOM-BC2) |\n| Pelto 2025 *Brief Bioinform* 26(2):bbaf130 | cross-study replicability | elementary BH-corrected methods most replicable; ANCOM-BC2 worst |\n\nReport the disagreement AS the result; verify current best practice against the latest tool docs rather than hard-coding one method.\n\n## Tool Taxonomy\n\n| Tool | Citation | Mechanism \u002F role | When |\n|------|----------|------------------|------|\n| ALDEx2 | Fernandes 2014 *Microbiome* 2:15 | Dirichlet Monte-Carlo posterior + CLR; tests each draw; reports expected effect + BH-adjusted p | conservative two-group anchor; small-to-moderate n |\n| ANCOM-BC2 | Lin & Peddada 2024 *Nat Methods* 21:83 | estimates per-sample sampling fraction and bias-corrects; structural zeros; pseudo-count sensitivity (`passed_ss`) | interpretable LFC + CI; covariates; multi-group |\n| MaAsLin2 | Mallick 2021 *PLoS Comput Biol* 17:e1009442 | general (mixed) linear model on transformed abundance | multivariable \u002F longitudinal \u002F metadata-rich |\n| MaAsLin3 | Nickols 2026 *Nat Methods* 23:554 | splits abundance (level when present) from prevalence (present\u002Fabsent); absolute-abundance mode | prevalence-vs-abundance separation; load data available |\n| LinDA | Zhou 2022 *Genome Biol* 23:95 | CLR regression with mode-based bias correction; asymptotic FDR | large cohorts; fast; native mixed model |\n| ZicoSeq | Yang & Chen 2022 *Microbiome* 10:130 | reference-taxa normalization + permutation FDR; winsorization | covariates; non-parametric permutation p; strong FP control |\n| LEfSe | Segata 2011 *Genome Biol* 12:R60 | Kruskal-Wallis + LDA effect size | exploratory biomarker discovery; NOT a formal FDR-controlled test |\n| DESeq2 | Love 2014 *Genome Biol* 15:550 | RNA-seq median-of-ratios size factor | caveat only; geometric-mean reference dies on sparse zero-heavy tables |\n\n## Decision Tree by Scenario\n\n| Scenario | Recommended | Why |\n|----------|-------------|-----|\n| Two groups, want a trustworthy conservative anchor | ALDEx2 | Dirichlet-MC + CLR; most reproducible\u002Fconservative (Nearing); gate on effect size |\n| Need interpretable LFC + CI, structural zeros, multi-group | ANCOM-BC2 | models and corrects per-sample sampling fraction; global\u002Fpairwise\u002FDunnett\u002Ftrend; `passed_ss` |\n| Large cohort, covariates, speed, mixed model | LinDA | CLR regression + bias mode; asymptotic FDR; fast; random effects in the formula |\n| Covariates + permutation-grounded non-parametric p | ZicoSeq | reference-taxa frame + permutation FDR |\n| Longitudinal \u002F many covariates \u002F flexible GLM | MaAsLin2 | `fixed_effects` + `random_effects`; normalization\u002Ftransform menu |\n| Prevalence-vs-abundance separation or absolute abundance | MaAsLin3 | logistic prevalence model + abundance model; load-data hook |\n| Inside a QIIME2 CLI pipeline | qiime composition ancombc + tabulate\u002Fda-barplot | native artifact flow (v1 ANCOM-BC; go to R for v2 `passed_ss`\u002Fmulti-group) -> qiime2-workflow |\n| Repeated \u002F paired samples | any tool above WITH a random effect | ignoring subject structure is pseudo-replication |\n| ALWAYS | run >=2 of the above, report the consensus | tool choice drives the hit list more than biology (Nearing 2022) |\n| Shotgun species table, not amplicon | -> metagenomics\u002Fmetagenome-visualization | same CoDA theory; different upstream pipeline |\n| Uncorrected t-test\u002FWilcoxon on TSS proportions | DO NOT | closure biases the test and there is no FDR control |\n\n## Filter Before Testing (a modeling knob, not housekeeping)\n\n**Goal:** Drop rare features before testing so the BH denominator is not crushed and log\u002FCLR transforms are well-behaved.\n\n**Approach:** Keep features present in at least 10-25% of samples (and optionally a mean-abundance floor); declare the threshold and confirm the headline result is not knife-edge-sensitive to it. Every tool exposes this (`prv_cut`, `min_prevalence`, `prev.filter`).\n\n```r\nlibrary(phyloseq)\nps \u003C- readRDS('phyloseq_object.rds')\n# prv_cut 0.10: a feature must appear in >= 10% of samples; raising to 0.25 removes more tests\n# (smaller BH correction, more power on survivors) but discards rare-but-real taxa - a declared choice\nkeep \u003C- filter_taxa(ps, function(x) sum(x > 0) >= 0.10 * nsamples(ps), TRUE)\n```\n\n## ALDEx2: The Conservative Floor of the Consensus\n\n**Goal:** Identify taxa that differ between two groups while propagating the sampling uncertainty of low-count features.\n\n**Approach:** Draw `mc.samples` Monte-Carlo instances from a Dirichlet posterior of the counts (this IS the zero handling - no explicit pseudocount), CLR-transform each instance against the geometric mean of all features (`denom='all'`), run the test on every draw, and report the EXPECTED effect size and BH-adjusted p over the draws.\n\n```r\nlibrary(ALDEx2)\ncounts \u003C- as.matrix(otu_table(ps))            # integer counts, taxa in ROWS\nif (!taxa_are_rows(ps)) counts \u003C- t(counts)\ngroups \u003C- as.character(sample_data(ps)$Group)\n\n# mc.samples 128: standard Monte-Carlo draws; 256+ for publication (more stable expected p)\nres \u003C- aldex(counts, groups, mc.samples = 128, test = 't', effect = TRUE, denom = 'all')\n# we.eBH = Welch expected BH-adjusted p (report this, NOT we.ep); wi.eBH = Wilcoxon equivalent\n# effect = median standardized effect = median(diff.btw \u002F max(diff.win)); the primary decision variable\nhits \u003C- res[res$we.eBH \u003C 0.05 & abs(res$effect) > 1, ]   # q AND effect floor (Gloor: gate on effect, not p alone)\n```\n\nGate on effect size AND q, not p alone: with large n trivially small CLR differences become \"significant,\" and Gloor's own guidance is that `|effect| > 1` is a strong ~2-SD signal. For >2 groups use `aldex.kw()`; for covariates the `aldex.glm()` + model.matrix route works but ALDEx2 is weakest here - prefer ANCOM-BC2\u002FLinDA\u002FMaAsLin2 for serious covariate or random-effect modeling.\n\n## ANCOM-BC2: Bias-Corrected LFC With a Sensitivity Safeguard\n\n**Goal:** Estimate an interpretable bias-corrected log-fold-change per taxon, with covariate adjustment, structural-zero handling, and a flag for hits that are hostage to the pseudo-count.\n\n**Approach:** Model log(observed count) as a function of covariates, estimate each sample's log sampling fraction as an offset and subtract it, then refit across a range of pseudo-counts and record how often each q-value flips (`passed_ss`).\n\n```r\nlibrary(ANCOMBC)\nout \u003C- ancombc2(data = ps, fix_formula = 'Group + Age + Sex',\n                rand_formula = NULL,        # '(1 | SubjectID)' for repeated measures - see Failure Modes\n                p_adj_method = 'BH',        # DEFAULT is 'holm'; set 'BH' deliberately for FDR\n                prv_cut = 0.10, lib_cut = 1000,\n                group = 'Group', struc_zero = TRUE, pseudo_sens = TRUE,\n                global = FALSE, pairwise = FALSE, n_cl = 2)\nres \u003C- out$res\n# a confident hit is BOTH significant AND robust to the pseudo-count. ANCOM-BC2 suffixes the\n# diff_\u002Fpassed_ss_ columns with the literal model-matrix coefficient (variable + factor level,\n# verbatim case, e.g. 'Grouptreated') - match it by pattern rather than hard-coding the case.\ndcol \u003C- grep('^diff_Group', names(res), value = TRUE)[1]\nrobust \u003C- res[res[[dcol]] & res[[sub('^diff_', 'passed_ss_', dcol)]], ]\n```\n\n`passed_ss` is the most valuable ANCOM-BC2-specific feature: a CLR\u002Flog model on sparse data is hostage to the zero-replacement constant, and `passed_ss` quantifies that per taxon. A hit with `passed_ss == FALSE` depends on the arbitrary pseudo-count - do not report it as confident. For >2 groups set `global=TRUE` (omnibus), `pairwise=TRUE` (mdFDR-controlled pairs), `dunnet=TRUE`, or `trend=TRUE`; results land in `out$res_global`\u002F`res_pair`\u002F`res_dunn`\u002F`res_trend`.\n\n## LinDA: Fast CLR Regression With Native Mixed Models\n\n**Goal:** Get FDR-controlled log2-fold-changes on a large cohort, including repeated-measures designs, without Monte-Carlo or EM cost.\n\n**Approach:** Fit ordinary linear regression on the CLR-transformed table covariate by covariate, estimate the compositional bias as the mode of the per-feature coefficients and subtract it; a random effect in the formula makes it a linear mixed model.\n\n```r\nlibrary(MicrobiomeStat)\notu \u003C- as.data.frame(otu_table(ps)); if (!taxa_are_rows(ps)) otu \u003C- t(otu)\nmeta \u003C- as.data.frame(sample_data(ps))\nfit \u003C- linda(feature.dat = otu, meta.dat = meta,\n             formula = '~ Group + Age + (1 | SubjectID)',   # random effect -> mixed model\n             feature.dat.type = 'count', prev.filter = 0.10, alpha = 0.05)\nfit$output[[1]]   # names(fit$output) are the model-matrix coefficient columns (e.g. 'Grouptreated' - the factor level keeps its case); per-feature: log2FoldChange, lfcSE, stat, pvalue, padj, reject\n```\n\nLinDA is the natural fast modern entry in a consensus panel and the cleanest route to mixed models. Yang & Chen rate it among the best FDR-power trade-offs.\n\n## MaAsLin2 \u002F MaAsLin3 and ZicoSeq (the rest of the panel)\n\n**Goal:** Fit covariate-rich or longitudinal differential-abundance models, or add a permutation-based panel member, when ALDEx2\u002FANCOM-BC2\u002FLinDA do not cover the design.\n\n**Approach:** Use MaAsLin2\u002F3 for multivariable GLMs with random effects, or ZicoSeq for a non-parametric permutation-FDR test against empirically selected reference taxa.\n\nMaAsLin2 fits a flexible per-feature GLM; its package DEFAULT is TSS + LOG + LM (not CLR), and `random_effects` is the canonical route for longitudinal designs. NOTE the orientation gotcha: it expects features in COLUMNS, samples in rows.\n\n```r\nlibrary(Maaslin2)\nfit \u003C- Maaslin2(input_data = as.data.frame(t(otu)), input_metadata = meta,\n                output = 'maaslin2_out', fixed_effects = c('Group', 'Age'),\n                random_effects = c('SubjectID'),\n                normalization = 'TSS', transform = 'LOG', analysis_method = 'LM',\n                min_prevalence = 0.10, max_significance = 0.05)\n# writes all_results.tsv \u002F significant_results.tsv with columns feature, metadata, coef, pval, qval\n```\n\nMaAsLin3 (`maaslin3()`) splits each feature into an abundance model (level when present) and a logistic prevalence model (present\u002Fabsent) tested jointly, and can ingest total-load measurements for absolute-abundance inference. ZicoSeq (`GUniFrac::ZicoSeq()`) winsorizes, posterior-samples, normalizes against empirically selected reference taxa, and returns permutation FDR (`zc$p.adj.fdr`) - a non-parametric panel member that accepts covariates via `adj.name`.\n\n## Consensus: Intersect the Tools\n\n**Goal:** Convert two or more per-tool hit sets into a confidence-graded result instead of one tool's answer.\n\n**Approach:** Collect the significant feature SETS (BH within each tool), then report the intersection as high-confidence, the union as exploratory, and tabulate, per taxon, how many of N tools agree and which ones. Never pool p-values across tools.\n\n```r\nsig_aldex \u003C- rownames(res)[res$we.eBH \u003C 0.05 & abs(res$effect) > 1]\nsig_linda \u003C- rownames(fit$output[[1]])[fit$output[[1]]$reject]   # [[1]] = the group coefficient (named 'Grouptreated')\nconfident  \u003C- intersect(sig_aldex, sig_linda)   # high-confidence\nexploratory \u003C- union(sig_aldex, sig_linda)       # report with the tool that found each\n```\n\n## Per-Method Failure Modes\n\n### Cherry-picking the tool with the prettiest result\n**Trigger:** running several tools and reporting only the one(s) that flag the favored taxon. **Mechanism:** that is uncorrected multiplicity hidden in the method menu (p-hacking by software). **Symptom:** \"the recommended method found X\" with no mention of the tools that disagreed. **Fix:** decide the panel a priori, report ALL tools, intersect for confident hits, disclose disagreement.\n\n### Uncorrected Wilcoxon\u002Ft-test on relative abundances\n**Trigger:** a per-taxon Wilcoxon\u002Ft-test on TSS proportions with no FDR correction. **Mechanism:** closure makes a naive test call every taxon \"decreased\" when one blooms, and hundreds of uncorrected tests inflate false positives. **Symptom:** dozens of \"significant\" taxa, all in the same direction, no q-values. **Fix:** use a CoDA\u002Freference-frame tool; if a simple test is used, BH-correct it and label the comparison as relative (Pelto 2025).\n\n### Pseudo-replication of repeated measures\n**Trigger:** longitudinal\u002Fpaired samples treated as independent rows. **Mechanism:** fewer independent units than rows inflates significance. **Symptom:** implausibly small p-values on a small subject count. **Fix:** a random effect - ANCOM-BC2 `rand_formula='(1|SubjectID)'`, MaAsLin2 `random_effects='SubjectID'`, LinDA `(1|SubjectID)` in the formula. Cross-check ANCOM-BC2 mixed-model output against LinDA\u002FMaAsLin2 (GitHub issue #111 reported `rand_formula` correctness problems in some versions).\n\n### Prevalence filter set blindly\n**Trigger:** an undeclared prevalence cut, or none at all. **Mechanism:** the cut decides which taxa are even tested and thus the BH landscape - it is a modeling choice. **Symptom:** the hit list changes materially between `prv_cut=0.1` and `0.25`. **Fix:** declare and justify the threshold; confirm the headline result survives moving it.\n\n### Relative change reported as absolute\n**Trigger:** \"taxon X doubled\" from a closed table with no load data. **Mechanism:** one taxon blooming compresses every other proportion. **Symptom:** whole-community \"depletion\" that is really one taxon rising. **Fix:** anchor to load (spike-in\u002Fflow\u002FqPCR) or MaAsLin3 absolute mode; otherwise state the claim is relative.\n\n### DESeq2\u002FedgeR on a sparse 16S table\n**Trigger:** RNA-seq median-of-ratios \u002F TMM on a zero-heavy ASV table. **Mechanism:** the geometric-mean size-factor reference collapses on zeros and the \"most features unchanged\" assumption is violated. **Symptom:** degenerate size factors, errors, or inflated hit counts that disagree with CoDA tools (Nearing). **Fix:** use a compositional tool; if DESeq2 is unavoidable, the `poscounts` estimator is the minimum mitigation - present as a caveat, not a recipe.\n\n### ANCOM-BC2 hit held hostage by the pseudo-count\n**Trigger:** reporting `diff_* == TRUE` without checking `passed_ss_*`. **Mechanism:** significance depends on the arbitrary zero-replacement constant. **Symptom:** a hit that vanishes when the pseudo-count changes. **Fix:** require `diff_* & passed_ss_*` for a confident call.\n\n## Quantitative Thresholds\n\n| Threshold | Source | Rationale |\n|-----------|--------|-----------|\n| Prevalence cut 10-25% (`prv_cut`\u002F`min_prevalence`\u002F`prev.filter`) | Nearing 2022; tool defaults (0.10) | rare features carry little information and crush the BH denominator; declare the value and test sensitivity |\n| BH q \u003C= 0.05 across taxa, within each tool | Benjamini-Hochberg 1995 *JRSS B* 57:289 | hundreds-thousands of features make uncorrected p meaningless; do not pool p across tools |\n| ALDEx2 `|effect| > 1` (with q \u003C= 0.05) | Gloor 2016 *J Comput Graph Stat* 25:971 | effect is a standardized median-ratio; ~2 SD is a strong signal; large n makes trivial diffs \"significant\" |\n| ALDEx2 `mc.samples` = 128 (256+ for publication) | Fernandes 2014 *Microbiome* 2:15 | Monte-Carlo draws; more draws stabilize the expected p |\n| ANCOM-BC2 `passed_ss == TRUE` required | Lin & Peddada 2024 *Nat Methods* 21:83 | flags hits whose significance is hostage to the pseudo-count |\n| Consensus of >=2 compositionally-aware tools | Nearing 2022 *Nat Commun* 13:342 | tool choice drives the hit list more than biology; intersection = confident |\n| ZicoSeq permutations `perm.no` >= 99 | Yang & Chen 2022 *Microbiome* 10:130 | permutation FDR resolution; raise for finer tail p |\n\n## Common Errors\n\n| Error \u002F symptom | Cause | Solution |\n|-----------------|-------|----------|\n| ALDEx2 returns NA effects \u002F errors | proportions or non-integer matrix passed | feed integer COUNTS with taxa in rows |\n| `passed_ss` column missing | `pseudo_sens = FALSE` | set `pseudo_sens = TRUE` (the default) |\n| Far fewer hits than expected | ANCOM-BC2 `p_adj_method` left at `holm` | set `p_adj_method = 'BH'` deliberately if FDR is wanted |\n| MaAsLin2 finds nothing \u002F orientation error | features in rows, not columns | transpose so samples are rows, features columns |\n| Mixed-model hits disagree across tools | `rand_formula` correctness varies by version | cross-check ANCOM-BC2 against LinDA\u002FMaAsLin2 |\n| Tools disagree on the hit list | normal - tool choice drives results | report the consensus and the disagreement, do not cherry-pick |\n| Many \"depleted\" taxa in a host\u002Fplant sample | host mitochondria\u002Fchloroplast 16S inflates the table | filter Mitochondria\u002FChloroplast features (see taxonomy-assignment) before DA |\n| Contaminant ASVs among the hits (low-biomass) | reagent kitome not removed before DA | run decontam upstream with negative controls (amplicon-processing; metagenomics\u002Fcontamination-controls) |\n\n## References\n\n- Fernandes AD, Reid JNS, Macklaim JM, McMurrough TA, Edgell DR, Gloor GB. 2014. Unifying the analysis of high-throughput sequencing datasets: characterizing RNA-seq, 16S rRNA gene sequencing and selective growth experiments by compositional data analysis. *Microbiome* 2:15.\n- Gloor GB, Macklaim JM, Fernandes AD. 2016. Displaying variation in large datasets: plotting a visual summary of effect sizes. *J Comput Graph Stat* 25:971-979.\n- Lin H, Peddada SD. 2020. Analysis of compositions of microbiomes with bias correction (ANCOM-BC). *Nat Commun* 11:3514.\n- Lin H, Peddada SD. 2024. Multigroup analysis of compositions of microbiomes with covariate adjustments and repeated measures (ANCOM-BC2). *Nat Methods* 21:83-91.\n- Mallick H, Rahnavard A, McIver LJ, et al. 2021. Multivariable association discovery in population-scale meta-omics studies. *PLoS Comput Biol* 17:e1009442.\n- Nickols WA, Kuntz T, Shen J, et al. 2026. MaAsLin 3: refining and extending generalized multivariable linear models for meta-omic association discovery. *Nat Methods* 23:554-564.\n- Zhou H, He K, Chen J, Zhang X. 2022. LinDA: linear models for differential abundance analysis of microbiome compositional data. *Genome Biol* 23:95.\n- Yang L, Chen J. 2022. A comprehensive evaluation of microbial differential abundance analysis methods: current status and potential solutions. *Microbiome* 10:130.\n- Yang L, Chen J. 2023. Benchmarking differential abundance analysis methods for correlated microbiome sequencing data. *Brief Bioinform* 24:bbac607.\n- Pelto J, Auranen K, Kujala JV, Lahti L. 2025. Elementary methods provide more replicable results in microbial differential abundance analysis. *Brief Bioinform* 26(2):bbaf130.\n- Nearing JT, Douglas GM, Hayes MG, et al. 2022. Microbiome differential abundance methods produce different results across 38 datasets. *Nat Commun* 13:342.\n- Segata N, Izard J, Waldron L, et al. 2011. Metagenomic biomarker discovery and explanation. *Genome Biol* 12:R60.\n- Love MI, Huber W, Anders S. 2014. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. *Genome Biol* 15:550.\n- Benjamini Y, Hochberg Y. 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. *J R Stat Soc Series B* 57:289-300.\n\n## Related Skills\n\n- diversity-analysis - Whole-community alpha\u002Fbeta\u002FPERMANOVA; answer \"do the communities differ\" before \"which taxa differ\"\n- taxonomy-assignment - Collapse ASVs to genus\u002Fspecies before per-taxon testing\n- amplicon-processing - Produces the ASV feature table tested here\n- qiime2-workflow - The qiime composition ancombc CLI route\n- metagenomics\u002Fabundance-estimation - Shared compositional\u002Fclosure\u002FCLR\u002Fzero\u002Fload-anchor theory\n- metagenomics\u002Fmetagenome-visualization - The same DA mechanics on shotgun profiler tables\n- experimental-design\u002Fmultiple-testing - FDR control and multiplicity across taxa\n",{"name":5,"source":24,"license":15,"tool_type":25,"description":7,"imported_at":26,"source_path":27,"primary_tool":28,"source_repository":29},"https:\u002F\u002Fgithub.com\u002FGPTomics\u002FbioSkills","r","2026-09-20","microbiome\u002Fdifferential-abundance\u002FSKILL.md","ALDEx2","GPTomics\u002FbioSkills","main","2d1750717747097b602acf30bd09a7cbd4b4c6ea","awesome_skills\u002Fbiology\u002Fbio-microbiome-differential-abundance","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous","2026-11-09 11:45:36",{"source":8,"skillName":5,"commands":37,"prompt":47,"zipAvailable":48},[38,41,44],{"tool":39,"command":40},"npx","npx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-microbiome-differential-abundance",{"tool":42,"command":43},"bunx","bunx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-microbiome-differential-abundance",{"tool":45,"command":46},"pnpm","pnpm dlx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-microbiome-differential-abundance","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-microbiome-differential-abundance\u002FSKILL.md",true,[50,56],{"path":51,"resourceType":52,"resourceTypeDesc":53,"sizeBytes":54,"sourceHtmlUrl":55},"examples\u002Faldex2_analysis.R","OTHER","其他",4350,"https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-microbiome-differential-abundance\u002Fexamples\u002Faldex2_analysis.R",{"path":57,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":60,"sourceHtmlUrl":61},"usage-guide.md","REFERENCE","参考文档",6880,"https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-microbiome-differential-abundance\u002Fusage-guide.md",[63,71,78,85,92,99],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":70},"01M7FFFZPC2025JQQHPHDX154J","dei-assessment","Assess diversity, equity, and inclusion with metrics. TRIGGERS - Use when user needs help with dei-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-assessment\u002FSKILL.md",[],[],"2026-11-16 10:05:34",{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":70},"01M7FFFWN8H4WA7A6Y398QKB6J","dei-action-plan","Create DEI action plans with goals and initiatives. TRIGGERS - Use when user needs help with dei-action-plan related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-action-plan\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":70},"01M7FFFSHTM0QZRVZ2NRJTJ0PY","culture-fit-assessment","Design culture add assessment processes. TRIGGERS - Use when user needs help with culture-fit-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fculture-fit-assessment\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":70},"01M7FFFMA12RKEQA38KHJSXXPS","cowork-to-code-bridge","Use an already-installed, independently verified cowork-to-code bridge to run narrowly approved actions on the user's own macOS, Linux, or WSL2 machine through a local file queue.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcowork-to-code-bridge\u002FSKILL.md",[],[],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":70},"01M7FFFH1F6M8YRGW5V487GH1K","counter-offer-response","Draft counter-offer responses for retention. TRIGGERS - Use when user needs help with counter-offer-response related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcounter-offer-response\u002FSKILL.md",[],[],{"skillNo":100,"name":101,"nameZh":6,"description":102,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":103,"categories":104,"keywords":105,"publishedAt":70},"01M7FFFDGEHV8Y7J9N40HZXW52","continuous-feedback-system","Design continuous feedback systems. TRIGGERS - Use when user needs help with continuous-feedback-system related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcontinuous-feedback-system\u002FSKILL.md",[],[]]