Bench to Bedside Immunotherapy Development

Key Takeaways

• This case study follows the full immunotherapy development journey - from NGS antibody screening and in silico selection, through in vivo combination therapy analysis, target expression safety profiling, clinical trial monitoring, and prognostic biomarker discovery
• ML-based methods reduced antibody screening time by 53%, accelerated in vivo response analysis by 80%, and cut biomarker discovery time by 72% - with 11 clinically relevant biomarkers identified within a week
• All data shown is publicly available or synthetic and representative of real client programmes; all images are anonymised

Jump to section

Introduction

This use case showcases the real analytical work being done across bench to bedside immunotherapy development programmes. It utilises publicly available data from a variety of sources (see Table 1) manipulated to create a unified story of preclinical drug discovery and development through to clinical deployment.

We demonstrate how effective data analysis, visualisation, and cutting-edge machine learning techniques can significantly enhance drug discovery and the development of therapeutics with a high likelihood of success - from target identification through to patient stratification in clinical trials.

Novel Immuno-Oncology Antibody Discovery

Our client required help developing novel antibodies with strong target binding affinity and specificity. An NGS antibody screen was used, alongside ML in silico methods, to identify antibodies with strong binding potential. The Nextflow airrflow pipeline was used to process raw FASTQ files. This same pipeline infrastructure underpins our T cell response immune profiling workflow. Following screening, antibodies selected for binding potential were assessed via ELISA affinity assays to identify the strongest candidates for further preclinical development.

Investigate Target Expression

When developing therapeutics against novel targets, a critical challenge is determining possible side effects through understanding on-target and off-target expression. A 'Weaver Plot' tool integrates gene data from multiple studies in a single view, enabling researchers to understand expression across different tissues and disease indications - identifying targets with limited off-tumour effects. This supports target identification and validation workflows where thorough assessment of on-target and off-target expression is critical for a successful IND submission.

In Vivo Preclinical Development

In the preclinical stage, in vitro and in vivo data were analysed to determine that the novel immunotherapy worked best in a particular cancer type and in combination with two standard-of-care treatments. Multiple assays were used to determine potential mechanisms of action, including digital pathology of tumour samples and single-cell RNA-Seq. No-code data visualisations allowed rapid interrogation of immune cell population changes induced by treatment.

Early Clinical Trial Analysis

Another client had multiple assets in various clinical stages. They needed to compare novel therapies with standard of care, identify responders and non-responders using RECIST criteria, and monitor therapy safety. Clinical trial data was brought into the platform from multiple sources in real time and analysed using no-code applications. This is exactly the kind of patient stratification capability that separates data-driven programmes from those relying on manual reporting cycles.

Previously, this client had to request and pay per report to evaluate ongoing trials, which could take days to arrive. Real-time access to version-controlled trial data enabled immediate decision-making - a direct answer to the data integrity and reproducibility challenges that affect clinical trial analysis.

Development of Prognostic Biomarkers

During clinical development it became clear that prognostic biomarkers were needed to stratify patients and improve odds of success in later-stage trials. Tumour and blood samples collected before treatment generated omics datasets. Both traditional differential expression and ML methods were used - with the R&D team discovering clinically relevant biomarkers within a single day of the omics dataset reaching the platform. This collaborative research environment accelerated the biomarker discovery process by months.

PCA analysis was used to visualise the complex dataset and identify distinguishing features between responders and non-responders. An XGBoost classifier was then trained to distinguish the two groups. The approach mirrors the bulk RNA-Seq workflow applied to public datasets, now applied to proprietary clinical data within a governed environment.

Conclusion

This case study covers some of the data types and analytical methods frequently applied across bench to bedside immunotherapy development programmes. With purpose-built infrastructure for precision medicine, the full range of required data types - from NGS antibody screens through to clinical trial omics - can be managed, analysed, and reported in a single governed environment.

Access to data and the ability to analyse it with cutting-edge ML tools should no longer be a blocker to clinical development. Once candidate biomarkers are identified, the next step is a rigorous biomarker validation strategy. For the multi-omics dimension of this work, see also our MOFA+ autoimmune clinical trial case study.

Data Sources

Talk to our team

Sonrai acts as an extension of your team, combining scientific expertise, transparent and reproducible analysis, and a governed research environment to generate decision-ready evidence that can be revisited, reproduced, and built upon throughout the life of a programme.