Bioinformatics and data analysis
Turning complex data into interpretable biological evidence
We analyse and integrate clinical, molecular and omics data to identify patterns, test hypotheses and build robust, reproducible and communicable results.
From data to results you can interpret and use
Talk to the team
What data can help clarify
We start from the question and the study design to decide which comparisons, associations or models are appropriate, and which limits apply to their interpretation.
- Associations
Which variables relate to the phenotype, progression or outcome under study?
We evaluate relationships between clinical, molecular and experimental variables, adjusting the analysis to the design and distinguishing exploratory findings from associations that need confirmation.
Where it applies- Group comparison Differences between conditions, phenotypes or degrees of involvement
- Course and progression Variables associated with trajectories or clinical outcomes
- Response
Which factors help explain the response to an intervention or treatment?
We combine baseline and longitudinal measurements to study heterogeneity of response, within-individual changes and differences between groups or time points.
Where it applies- Interventions Biological change associated with treatment, nutrition or lifestyle
- Longitudinal follow-up Individual trajectories and change over time
- Stratification
Are there subgroups with different biological or clinical profiles?
We explore multivariate patterns and sources of heterogeneity to characterise profiles, without turning a visual separation or an exploratory model into a validated classification.
Where it applies- Complex phenotypes Characterisation of heterogeneous populations
- Signal prioritisation Candidate variables for a validation phase
- Integration
How do we connect clinical, molecular and omics data in a single analysis?
We combine layers generated by Biosfer with external data when doing so improves interpretation and there is a clear correspondence between samples, variables and objectives.
Where it applies- Multi-omics integration Metabolomics, biomarkers and other molecular layers
- Clinical–molecular context Connecting molecular signals with phenotype and outcomes
These contexts are examples. The analysis plan is defined by the design, the size and structure of the sample, the data quality and the level of validation each conclusion requires.
From data to an interpretable answer
We agree on the starting point and move forward through a traceable process, proportionate to the question and to the level of evidence the study needs.
-
Plan
Design the analysis
Together with the research team we define the question, the design and the statistical strategy before the analysis starts.
- Question, hypothesis and outcome variables
- Groups, measurement time points and planned comparisons
- Power, covariates and statistical plan
Cases and controls matched for age, sex and BMI, with serum metabolomics from the follow-up visit. -
Prepare
Prepare the data for analysis
When the project combines different sources, we agree how to structure, code and link them, preserving the provenance and traceability of each variable.
- Consistency, missing values and detection limits
- Pre-processing and normalisation where appropriate
- Linking clinical data, biomarkers and metabolomics
Sources are linked in a documented, versioned and reproducible analysis dataset. -
Analyse
Describe, compare and estimate
We estimate differences, associations or effects with their uncertainty, according to the design (cross-sectional, longitudinal or interventional) and the nature of the data.
- Between-group comparison and baseline-to-end change
- Longitudinal models and adjusted estimation
- Multivariate structure and heterogeneity of response
Effect size and its confidence interval take priority over significance alone. -
Model
Classify, predict or stratify
When the objective requires it, we assess how much metabolomic information adds beyond clinical variables.
- Clinical, metabolomic and combined models
- Discrimination, calibration and variable contribution
- Validation, sensitivity and generalisability
Comparing models shows the added information provided by metabolomics. -
Translate
Turn the results into a useful tool
If the project needs it, we turn the analysis into an environment ready for interpretation or application.
- Figures, tables and a reproducible methods section
- Dashboards and monitoring tools
- Apps, scientific software and SaaS solutions
A validated model can become a research or monitoring tool.
Before starting a data project
01 Can you analyse data not generated by Biosfer?
Yes. We can work with clinical, retrospective, longitudinal or omics datasets generated by other teams. Before starting we review formats, metadata, design, quality and the feasibility of the proposed question.
02 Can you integrate metabolomics with clinical variables or other omics?
Yes, provided the observations are comparable and there is enough information to link the different sources. The strategy depends on the scale, the quality and the correspondence between samples and variables.
03 Do you develop predictive models?
We can develop and evaluate models when they answer a defined question and the study size and design allow it. We clearly distinguish development, internal validation and independent validation before discussing generalisation.
04 Can you support analyses aimed at publication?
Yes. The scope can include analytical strategy, figures, tables, methods description, results and support with interpretation. Responsibilities and deliverables are agreed for each collaboration.
05 Do you work with longitudinal data or repeated measures?
Yes. We plan the analysis taking into account the dependence between measurements, the follow-up time points, individual variability and incomplete data.
06 Can you help identify candidate biomarkers?
Yes. We can explore and prioritise signals associated with a phenotype or outcome. Candidates obtained in an exploratory phase need a subsequent confirmation and validation strategy.
Plan your data analysis
Tell us what data you have, how it was generated and what the study needs to answer. We will propose the analysis plan, the deliverables and the appropriate level of validation.
Send data and question