Unilever SERS Collaboration: A Nextflow Pipeline for BIPODE Analysis of High Throughput Transcriptomics Data
There's nothing more satisfying for us at Seqera than seeing the real-world impact of the work we do with our customers. A recent paper in Computational Toxicology, the product of a collaboration between Seqera and Unilever's Safety, Environmental and Regulatory Science group, is a case in point.
The paper, "The BIPODE package for analysis of high throughput transcriptomics concentration–response data", describes a project undertaken by Seqera's Professional Services team in collaboration with scientists from the Unilever SERS team. The work builds on an algorithm developed at Unilever for estimating points of departure (PoDs): the concentration at which a compound first produces a measurable change in cells. PoDs are a quantitative measure of potency for use in next-generation risk assessment (NGRA), where in vitro and computational methods are used to assess the safety of consumer-product ingredients without animal testing.
BIPODE stands for Bayesian Inference for Point-of-Departure Estimation. What sets it apart is how it reads the shape of a concentration-response curve. Most methods fit the data to the closest match from a fixed set of curve shapes. BIPODE instead lets a Gaussian process shape the curve directly from the data, so a single model can capture almost any response. It also reports each result as a range with a confidence level, not a single number, which matters when analysis runs across thousands of genes at once, as in transcriptomics. BIPODE finds the lowest concentration affecting each gene, then combines these into a single figure for the whole experiment: the point below which nothing responds.
The Unilever SERS team already had Python code implementing the algorithm and an early workflow in place. Working with Seqera, they turned that early vision into a production-ready software, built to be shared and deployed at scale.
The Seqera team packaged the existing algorithm together with a test suite to guarantee its behaviour and stability. You can install it today from PyPI and Bioconda as bipode-httr. The Bayesian model itself runs on Stan, a well-established engine for this kind of inference, so anyone who installs the package gets the same results.
From there the team built a Nextflow workflow, bipode-nextflow, to nf-core community standards and best practices. Because the per-gene fits are independent of one another, the workflow can distribute them across HPC clusters or cloud infrastructure using Nextflow, with software dependencies packaged via Conda, Docker, or Apptainer.
Results are presented in an interactive report generated with MultiQC, allowing users (computational biologists, risk assessors) to explore the global PoD, drilling down into individual genes and identify outliers in one document, supporting multi-disciplinary team work and collaboration.
The publication uses three case studies to demonstrate that the new bipode-nextflow workflow produces reproducible global PoDs that compare favorably with the most credible existing methods.
“‘A key success of the BIPODE project has been enabling reproducible, independent implementation of the pipeline across our global collaborations. Allowing our partners to focus on science, rather than software installation and troubleshooting.’’
Mark Liddell, Bioinformatics Engineer, Unilever SERS team
Work like this is a reminder of the impact Seqera has with scientific teams across the world. In this case, helping to support the deployment at scale of next generation, non-animal safety approaches
About Unilever
Unilever is a global consumer goods company that has pioneered the use of alternatives to animal testing for decades, publishing over 600 scientific papers since the 1980s. Unilever does not test on animals and has committed to end animal testing globally. Next generation, non-animal safety approaches now exist that provide significant scientific and ethical benefits over animal tests. Unilever SERS team collaborate with over 70 partners globally to develop, evaluate and apply non-animal safety approaches.