OUR approach
Analysis starts with experimental context
Analysis done at a distance from the bench is missing the information that matters most.
Ours is done by people who know how the samples were run: how the plates were laid out, which markers sat near the limit of detection, where the batch boundaries fall, and what the QC flagged.
Without that context, analytical decisions drift away from the experiment that generated the data.
Interpretation is part of delivering the result.
WHY bioinformatics & Data analysis
What you gain
Talk to a scientist about your dataset
WHY bioinformatics & data analysis
What you receive
Processed data
QC-assessed, with normalisation documented and reproducible.
Raw data
Delivered alongside processed outputs, so your team can reproduce or extend the analysis independently.
Analysis report
Written methods, results and limitations.
Publication-quality figures
Prepared to journal requirements.
Methods text
Written so it can be adapted for manuscripts or regulatory submissions.
Scientific interpretation
A conversation about what the data means, with the scientists who generated it.
Included in every project
Scoped and quoted separately
HOW IT WORKS
From raw signal to biological interpretation
Steps 1 and 2 are standard on every project and arrive with your data. Steps 3 to 6 are the statistical and interpretive work. That is not included as standard: it is scoped and quoted separately.
01
Quality control and normalisation
Internal-control normalisation, inter-plate calibration, and detectability assessment against background.
Values below the limit of detection are handled by an approach documented and agreed in advance, so no substituted value travels through the downstream tests disguised as a measurement.
02
Batch and plate effects
Identified and corrected where correction is scientifically defensible.
Where a batch effect is confounded with the study arm, we flag it. Removing it mathematically would not solve it.
03
Differential abundance
Depending on study design: group comparisons, longitudinal and repeated-measures models, covariate adjustment and multiple-testing correction.
04
Structure and pattern
Dimensionality reduction, clustering and correlation analysis, which reveal phenotypes and biomarker modules invisible when every analyte is treated independently.
05
Pathway and biological interpretation
Enrichment analysis connecting significant biomarkers to biological processes and pathways.
06
Panel and classifier development
Where appropriate, feature selection and predictive modelling to support classifier development, with cross-validation and a clear account of what still requires confirmation in an independent cohort.
Talk to a scientist
Proteomics data analysis
Each platform needs its own analysis
NULISA, Simoa and MSD produce different data, with different QC behaviour and failure modes. Analysing them the same way is one of the commonest sources of misleading results.
NULISAseq
A sequencing-derived readout, processed as count data and reported in NPQ units. Treat it like conventional immunoassay output and you get answers that look reasonable and are wrong.
Simoa and MSD
Absolute concentrations against calibration curves. That changes how limits of quantification, dilution and cross-plate comparison are handled.
Across all three
We check which analytes were genuinely detected before any comparison is made, so no result rests on a value the assay never measured.
Sequencing, genomics and multi-omic analysis
Genomics analysis is built around your study, so it’s scoped and quoted per project.
Genomics pipelines are built, not bought
Metagenomics
Multi-omic integration
specialist matrices
Difficult samples need careful statistics
Unusual sample types challenge the analysis as well as the assay.
Bronchoalveolar lavage fluid
Dilution varies from procedure to procedure.
Nasal lining fluid
Samples are diluted into buffer, often by an unmeasured amount.
Exhaled breath condensate
There is no agreed standard for normalisation.
In each case, a difference between samples can come from how they were collected, not the biology.
Detection comes first. No correction can rescue a biomarker the assay never measured.
analysis & data
Design the analysis before generating the data
The cheapest analytical problem to fix is the one prevented at study design. Bring us in before sample collection.
1. Power and sample size
2. Plate layout and randomisation
3. Controls and calibrators
4. Panel selection
Which targets are realistically likely to clear the limit of detection in your intended matrix, because a panel full of undetectable analytes is expensive and uninformative.
5. Collection and handling
Sample collection and handling decisions made with downstream analytical requirements in mind.
Talk to us before your study runs
Custom pipelines and automated reporting
For programmes generating the same data repeatedly, we build the analytical pipeline once and apply it consistently thereafter. The analysis stops being a project each time and becomes infrastructure.
Automated reporting workflows turn batch sample data into participant-level outputs: marker values against reference ranges, composite scores, interpretation text, and secure report distribution, generated without anyone reformatting a spreadsheet.
If you are running the same panel repeatedly and reformatting results by hand each time, that workflow is a candidate for automation. The cost of building the pipeline is usually recovered within a few batches.
trust
Data handling and governance
Tell us about your study
Send us your study design, or a dataset you are stuck on, and a senior scientist will tell you what the data can and cannot support.