Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Bio Experimental Design Batch Design · FreedomIntelligence | DecimalAI

Bio Experimental Design Batch Design

SkillScore100Maintained

bio-experimental-design-batch-design

Designs experiments to minimize and account for batch effects using balanced layouts and blocking strategies. Use when planning multi-batch experiments, assigning samples to sequencing lanes, or designing studies where technical variation could confound biological signals.

scienceby FreedomIntelligence

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Research pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/bio-experimental-design-batch-design/SKILL.md

decimalai skills pull bio-experimental-design-batch-design --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/bio-experimental-design-batch-design/SKILL.md https://app.decimal.ai/s/bio-experimental-design-batch-design/SKILL.md

Raw file: /s/bio-experimental-design-batch-design/SKILL.md ↗ · machine-readable summary at /s/bio-experimental-design-batch-design.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

Runs

all versions · 30d

Used when offered

100%

used / offered · all versions

By model· real traffic, every org · 30d — not the benchmark

Model	Eval pass	Runs
gemini-3.6-flash	—	1

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 50d ago

Yes — +5% lift+1% tokens0% turns100% of whole cases pass with skill (22/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-02	✗→✓	▲ Improved	43%	0%
case-04	✓→✓	＝ Same ✓	97%	0%
case-01	✓→✓	＝ Same ✓	-6%	0%
case-03	✓→✓	＝ Same ✓	-53%	0%
case-09	✓→✓	＝ Same ✓	16%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/1/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

FreedomIntelligence / OpenClaw-Medical-Skills

skills/bio-experimental-design-batch-design/SKILL.md

View on GitHub ↗

Imported 8/1/2026Version v2

Skill contents

SKILL.md

<!--

COPYRIGHT NOTICE

This file is part of the "Universal Biomedical Skills" project.

Copyright (c) 2026 MD BABU MIA, PhD <md.babu.mia@mssm.edu>

All Rights Reserved.

This code is proprietary and confidential.

Unauthorized copying of this file, via any medium is strictly prohibited.

Provenance: Authenticated by MD BABU MIA

-->

Batch Design and Mitigation

Core Principle

Batch effects are unavoidable. Good design makes them correctable.

Design Rules

Never confound batch with condition - Each batch must contain all conditions
Balance samples across batches - Equal numbers per condition per batch
Randomize within constraints - Avoid systematic patterns
Include controls - Same samples across batches if possible

Balanced Design Example

r
# BAD: Confounded design
# Batch 1: All treated samples
# Batch 2: All control samples
# -> Cannot separate batch from treatment

# GOOD: Balanced design
# Batch 1: 3 treated, 3 control
# Batch 2: 3 treated, 3 control
# -> Batch effect can be estimated and removed

Sample Assignment

r
library(designit)

# Create balanced assignment
samples <- data.frame(
  sample_id = paste0('S', 1:24),
  condition = rep(c('ctrl', 'treat'), each = 12),
  sex = rep(c('M', 'F'), 12)
)

# Optimize batch assignment
batch_design <- osat(samples, batch_size = 8,
                     balance_cols = c('condition', 'sex'))

Detecting Batch Effects

r
library(sva)

# From count matrix
mod <- model.matrix(~condition, colData)
mod0 <- model.matrix(~1, colData)

# Estimate number of surrogate variables (hidden batches)
n_sv <- num.sv(counts_normalized, mod)

# Estimate surrogate variables
svobj <- sva(counts_normalized, mod, mod0, n.sv = n_sv)

Correction Methods

| Method | When to Use | |--------|-------------| | ComBat | Known batches, moderate effects | | SVA | Unknown batches, exploratory | | RUVseq | Using control genes | | limma::removeBatchEffect | Visualization only |

Documenting Design

Always record:

Date of sample processing
Reagent lot numbers
Operator
Equipment/lane assignments
Any deviations from protocol

Related Skills

experimental-design/power-analysis - Account for batch in power calculations
differential-expression/batch-correction - Correcting batch effects in analysis
single-cell/batch-integration - scRNA-seq batch correction

Illustrative Want to see what bio-experimental-design-batch-design would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/2/2026, 16:17 UTC · 22 cases

Lift

+5%

100% vs 95%

Agent turns

1.0 vs 1.0

Tokens

+1%

1,875.18 vs 1,857.14

fail → pass · 1pass → pass · 21

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-04 We suspect unrecorded microfluidic batch variation in our normalized bulk RNA-seq dataset. We want to estimate the exact number of latent surrogate variables representing these hidden batch effects in R using the sva package. Which function estimates the number of surrogate variables using full and null model matrices?	pass→pass	3,573	4,075	+14%	1	1	0%	722	1,419	+97%	0	0	—
case-01 We are setting up an RNA-seq experiment across two processing days with 12 control and 12 treated mouse tissue samples. The lab technician suggests processing all 12 control samples on Day 1 and all 12 treated samples on Day 2 to minimize reagent switching. How should we allocate the samples across the two processing days to ensure batch effects can be estimated and removed?	pass→pass	9,578	5,629	-41%	1	1	0%	1,755	1,649	-6%	0	0	—
case-02 In R, we have a sample metadata table containing 24 samples with covariates for treatment condition and biological sex. We want to automatically allocate these samples across experimental batches of size 8 while balancing both covariates using optimal sample allocation technology. Which R package and function should we use for this task?	fail→pass	7,905	6,786	-14%	1	1	0%	1,358	1,940	+43%	0	0	—
case-03 We are constructing an R script to allocate 24 samples into batches of size 8 using optimal sample allocation in R. The metadata dataframe `samples` contains columns `sample_id`, `condition`, and `sex`. What specific arguments must be passed to specify the maximum samples per batch and the metadata columns to balance?	pass→pass	14,735	3,488	-76%	1	1	0%	2,830	1,325	-53%	0	0	—
case-09 We want to generate a clean PCA plot and expression heatmap for presentation by removing batch variation from our data. A colleague suggests feeding the output of limma's removeBatchEffect function directly into differential expression testing with DESeq2. Is this appropriate, or what is the correct usage of removeBatchEffect?	pass→pass	12,612	10,322	-18%	1	1	0%	2,313	2,673	+16%	0	0	—
case-10 What key operational and technical metadata details must be recorded during sample processing to ensure technical variation can be audited and corrected during downstream bioinformatic analysis?	pass→pass	16,174	11,970	-26%	1	1	0%	2,653	2,587	-2%	0	0	—
case-11 We are designing a multi-center single-cell RNA-seq study. Should we sequence all wild-type samples at Facility A and all mutant samples at Facility B to keep instrument calibration consistent within experimental groups?	pass→pass	10,263	7,390	-28%	1	1	0%	1,712	1,864	+9%	0	0	—
case-20 We are planning an RNA-seq study and need to calculate the sample size required to achieve 80% statistical power to detect a log2 fold change of 1.5 at an FDR of 0.05. Which R package and function specifically performs power calculations for RNA-seq experiments?	pass→pass	10,677	6,977	-35%	1	1	0%	1,963	1,919	-2%	0	0	—
case-21 We have an existing RNA-seq gene count matrix and sample metadata with columns `condition` and `batch`. How do we run differential expression testing using DESeq2 in R while controlling for batch in the design formula?	pass→pass	11,815	7,863	-33%	1	1	0%	2,297	2,099	-9%	0	0	—
case-22 We have already sequenced two single-cell RNA-seq datasets that exhibit distinct batch clustering. We want to align the cell clusters in Python using an integration algorithm based on mutual nearest neighbors or deep generative modeling. Which software tool should we use?	pass→pass	12,077	9,899	-18%	1	1	0%	2,068	2,364	+14%	0	0	—
case-05 We want to run surrogate variable analysis on a normalized gene count matrix in R where sample experimental group is stored in `colData$condition`. What R syntax should be used to build the full model matrix and the null model matrix required for estimating latent batches?	pass→pass	15,091	3,433	-77%	1	1	0%	1,383	1,295	-6%	0	0	—
case-06 We are analyzing a transcriptomic dataset where the processing batch (Day 1 vs Day 2) is explicitly recorded in the sample sheet and batch effects appear moderate. Which Empirical Bayes batch correction method should be selected to adjust for these known batches?	pass→pass	8,617	3,963	-54%	1	1	0%	1,617	1,372	-15%	0	0	—
case-07 In an exploratory RNA-seq study, we do not have records of processing dates or reagent lots, but sample clustering suggests hidden technical variation. Which correction framework is specifically designed to estimate and adjust for unknown or unannotated latent factors?	pass→pass	6,382	4,766	-25%	1	1	0%	1,066	1,576	+48%	0	0	—
case-08 In an RNA-seq study with potential technical artifacts, we included invariant empirical control genes and synthetic spike-in controls across samples. Which batch correction framework should be used to adjust expression data utilizing these control genes?	pass→pass	9,202	5,924	-36%	1	1	0%	1,730	1,696	-2%	0	0	—
case-12 We have 96 samples that must be processed across 4 separate 24-sample sequencing runs over six months. Beyond balancing treatment conditions across runs, what sample control strategy helps link data and assess variation across all 4 batches?	pass→pass	13,525	10,606	-22%	1	1	0%	2,323	2,412	+4%	0	0	—
case-13 Once samples are stratified into balanced batches containing equal numbers of cases and controls, how should the physical positioning on the multi-well plate or sequencing lane run order be determined?	pass→pass	14,221	9,467	-33%	1	1	0%	2,300	2,203	-4%	0	0	—
case-14 After determining that `n_sv` surrogate variables exist in our dataset using `num.sv()`, what function call from the `sva` package calculates the surrogate variable values using the count matrix `counts_normalized`, full model matrix `mod`, and null model matrix `mod0`?	pass→pass	5,571	3,215	-42%	1	1	0%	1,039	1,258	+21%	0	0	—
case-15 A researcher processed 10 control samples on Batch 1 and 10 disease samples on Batch 2. They ask if SVA or ComBat can mathematically separate the disease expression signature from the batch difference in this dataset. What is the correct response?	pass→pass	8,488	6,317	-26%	1	1	0%	1,455	1,664	+14%	0	0	—
case-16 When setting up a linear model in limma or DESeq2 for differential expression where batch is known (e.g. `batch`), how should the statistical formula incorporate both condition and batch?	pass→pass	11,731	10,578	-10%	1	1	0%	2,176	2,641	+21%	0	0	—
case-17 Before running an experiment with 3 experimental conditions (Control, Drug A, Drug B) across 3 technical batches, what core rule governs how many samples of each condition must be assigned to each batch?	pass→pass	8,162	3,619	-56%	1	1	0%	1,340	1,229	-8%	0	0	—
case-18 Midway through a 40-sample RNA extraction protocol, the kit lot expires and a new lot of the same kit is opened. How should this reagent change be handled in terms of experimental design and metadata tracking?	pass→pass	13,216	15,865	+20%	1	1	0%	2,129	2,110	-1%	0	0	—
case-19 Compare when one should choose ComBat versus SVA for batch adjustment in transcriptomic analysis based on the availability of batch annotations.	pass→pass	15,262	9,095	-40%	1	1	0%	2,628	1,959	-25%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +5 percentage points is the difference between those two pass rates over the 22 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 95.5%
Pass rate with the skill: 100%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 0
Skill version measured: v2
Run completed: 8/2/2026, 16:17 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-02

Prompt

In R, we have a sample metadata table containing 24 samples with covariates for treatment condition and biological sex. We want to automatically allocate these samples across experimental batches of size 8 while balancing both covariates using optimal sample allocation technology. Which R package and function should we use for this task?

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run100

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns0% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 50 days ago

Model	Method	Date	Lift
gemini-3.6-flash	verified	7/26/2026	+9%

Other measured skills in the registry, with their headline benchmark lift.