Experimental design & power
Study design, test selection, and power and sample-size calculations before you generate data—so the experiment can answer the question.
Core Bioinformatics
Biostatistics & Visualization
From raw tables to rigorous answers and clear figures. We design the analysis, fit the right models—regression, mixed-effects, survival, machine learning—and turn the results into publication-quality visuals and interactive dashboards, all reproducible from documented, version-locked code.
Any tabular data in
Models · figures · dashboards out
Typical turnaround 3–7 days
Individual pricing from $149
Illustrative sample output
Group A
Group B
censored
Overview
Rigorous statistics & clear figures, without the in-house overhead
Good statistics is where a study earns its conclusions—the right design, the right model for the data, honest handling of multiple comparisons, and figures that show the result plainly. The hard part is rarely running a test; it is choosing a method whose assumptions match your data, accounting for the structure of the experiment, and presenting the outcome so reviewers and readers trust it.
We build and run that analysis for you. Whether you need a power calculation before you start, a mixed-effects model for a longitudinal study, a survival analysis, a machine-learning classifier, or a set of publication-ready figures, we apply established, well-documented methods—and return results with the code and every package version recorded so the analysis reproduces exactly.
Capabilities
What we analyse
One service spanning the analysis lifecycle—from design and power through modeling and machine learning to figures and dashboards.
Statistical modeling
Linear, generalized-linear, and mixed-effects models for grouped, longitudinal, and hierarchical data, with careful assumption checks.
Differential & multiple testing
Differential-expression and group-comparison statistics with proper multiple-testing correction and effect-size reporting.
Survival analysis
Kaplan-Meier estimates, log-rank tests, and Cox proportional-hazards models—with competing-risks methods where needed.
Multivariate & clustering
Dimensionality reduction and unsupervised structure—PCA, UMAP, and clustering to reveal patterns and subgroups.
Machine learning
Classification and regression with proper cross-validation, feature selection, and honest performance evaluation—no leakage.
Publication-quality figures
Clear, journal-ready figures—volcano plots, heatmaps, forest plots, survival curves—built to each journal's specifications.
Dashboards & reports
Interactive dashboards and reproducible reports that let collaborators explore results—rendered from documented code.
The Pipeline
How a statistical analysis runs
A transparent, best-practice sequence—methods chosen for your data and pre-specified where possible, and every step documented. Nothing is a black box.
Steps are adapted to your project: design help vs. analysis of existing data, a single test vs. a full model, static figures vs. an interactive dashboard. We agree the analysis plan with you before running it.
Scope & design
We define the question, choose the right test or model, and—if you are still planning—run power and sample-size calculations.
Tools: pwr · simr · analysis plan
Data QC & wrangling
Data is cleaned, reshaped, and validated—missingness, outliers, and coding checked before any modeling.
Tools: tidyverse · data.table · janitor
Exploratory analysis
Distributions, summaries, and first plots to understand the data and check the assumptions the model will rely on.
Tools: ggplot2 · skimr · GGally
Modeling & testing
The chosen model is fit—regression, mixed-effects, survival, or machine learning—with diagnostics and assumption checks.
Tools: lme4 · survival · tidymodels
Correction & validation
Multiple-testing correction, cross-validation, and sensitivity checks so results are robust, not artifacts of chance or overfitting.
Methods: BH-FDR · resampling · sensitivity
Visualization
Results become clear, journal-ready figures—or an interactive dashboard for exploration and sharing.
Tools: ggplot2 · ComplexHeatmap · Shiny
Reproducible reporting
A written methods and results summary, the figures, and version-locked code so the whole analysis reproduces exactly.
Tools: Quarto · R Markdown · renv
Tools & Technologies
Established, peer-reviewed tools—matched to your data
We select from the field's standard toolkit rather than forcing every dataset through one pipeline. A representative set of what we work with:
R & tidyverse
R
tidyverse
dplyr
data.table
Bioconductor
Statistical modeling
lme4
emmeans
glmmTMB
nlme
statsmodels
brms
Survival analysis
survival
survminer
cmprsk
mlr3proba
Machine learning
tidymodels
caret
scikit-learn
glmnet
xgboost
ranger
Multivariate & reduction
PCA
UMAP
t-SNE
factoextra
vegan
Visualization
ggplot2
ComplexHeatmap
patchwork
matplotlib
seaborn
Dashboards & interactivity
Shiny
Plotly
flexdashboard
Dash
Reproducible reporting
Quarto
R Markdown
knitr
Jupyter
renv
Foundations
Standards & ecosystems we build on
Sound analysis rests on established tools and reporting standards. We work within the community's trusted foundations.
- R & CRAN
- The statistical-computing language and its vetted package archive.
- Bioconductor
- Peer-reviewed packages for the statistics of genomic data.
- tidyverse & tidymodels
- Consistent, readable frameworks for data analysis and modeling.
- Python scientific stack
- NumPy, pandas, SciPy, statsmodels, and scikit-learn.
- Stan
- Probabilistic programming for Bayesian modeling where it fits.
- CONSORT & STROBE
- Reporting guidelines for trials and observational studies.
- Benjamini–Hochberg
- False-discovery-rate control for high-dimensional testing.
- FAIR principles
- Findable, accessible, interoperable, and reusable data practice.
- Quarto & renv
- Reproducible reporting and version-locked project environments.
Choosing an Approach
Classical tests vs. regression vs. machine learning
Different questions call for different tools. A quick orientation; we will help you pick the right one for your data.
| Dimension | Classical tests | Regression models | Machine learning |
|---|---|---|---|
| Answers | Is there a difference? | How do variables relate? | Can we predict the outcome? |
| Assumptions | Stronger, well-defined | Explicit & checkable | Fewer; data-driven |
| Typical output | p-value & effect size | Coefficients & intervals | Predictions & importance |
| Interpretability | High | High | Varies; needs care |
| Best suited to | Simple, focused comparisons | Inference & adjustment | Prediction & many features |
What You Receive
A complete, documented deliverable
Not just a p-value in an email—every output you need to understand, publish, and reproduce the analysis.
- Cleaned, analysis-ready dataset with a documented codebook
- Model results with effect sizes & confidence intervals (TSV / XLSX)
- Assumption checks & diagnostics
- Publication-quality figures (vector PDF / SVG / PNG)
- Interactive dashboard or app, where applicable
- A written methods & results summary for your manuscript
- Reproducible code with every package version recorded
Built for reproducibility, not just a result
Every analysis follows documented best practices—an analysis plan agreed up front, methods whose assumptions we check, multiple-testing correction, and honest reporting of uncertainty—so a result is earned rather than fished for. We separate pre-specified analyses from exploratory ones and say which is which. Each project ships with version-locked code and environment files so it reproduces exactly.
The practical payoff: your methods section writes itself, a reviewer can re-run the analysis, and a result from today can be reproduced a year from now. We will also tell you honestly when a design or sample size won't support the conclusion you're after.
FAQ
Biostatistics & visualization questions
What researchers and project leads most often ask before starting an analysis project.
Almost anything tabular — CSV, Excel, database exports, or the output tables from an omics pipeline. We clean and reshape the data, then model and visualise it. If you have a study design and a question, that is enough to start.
Yes. Before you generate data, we help define the design, choose the right statistical test or model, and run power and sample-size calculations so the study is adequately powered to answer your question.
Yes. We handle Kaplan-Meier estimates and Cox proportional-hazards models for survival data, and linear and generalized linear mixed models for repeated-measures, longitudinal, and hierarchical designs.
Yes. Alongside publication-quality static figures, we build interactive dashboards and apps (for example with R Shiny or Plotly) so you and your collaborators can explore results directly.
Yes. We produce publication-ready figures and a clear methods and results write-up, and can run the additional analyses reviewers request — with the code and versions recorded so everything is reproducible.
Cleaned data, model results with effect sizes and confidence intervals, publication-quality figures or an interactive dashboard, a written methods and results summary, and reproducible code with every package version recorded.
Related
Where statistics supports the work
Have data that needs the right statistics?
Tell us your organism, data type, and question—we'll scope it honestly, including if a different design would serve you better.