Customers › Population genomics
Case study · Population genomics
27% of breast cancers had treatment-relevant findings
A University of Cambridge–led study in The Lancet Oncology analyzed whole genomes from 2,445 breast cancers in the Genomics England Research Environment. Three Lifebit scientists are co-authors. The data never leaves the source.
01 · The challenge
Turn whole genomes into treatment decisions
Whole-genome sequencing produces large, complex data per patient. Linking it to clinical and mortality records at scale needs long, multi-stage pipelines and heavy compute, inside a governed environment.
Large data volumes per genome
Complex, multi-stage clinico-genomic pipelines
Thousands of tumor-normal genome pairs to profile, linked to outcomes
02 · What we deployed
Analysis inside the Genomics England research environment
Researchers worked in the Genomics England Research Environment, where Lifebit’s platform provides cloud-scale analysis. Lifebit scientists Vladyslav Dembrovskyi, Rosie Davies, and Maria Chatzou Dunford are co-authors and contributed resources. No patient-level data left custody, and any approved researcher can reach the same data “in the same manner that the authors accessed the data.”
Data stays in custody
Whole-genome and linked mortality data were analyzed in place in the National Genomic Research Library.
Genome-wide profiling at scale
Driver mutations, mutational signatures, HRD, mismatch repair deficiency, and tumor mutational burden for every tumor.
One governed dataset
Researchers from Cambridge, Genomics England, Lifebit, and NHS and European partners worked on the same governed data.
How it works
From connected data to approved results
Tumor and matched normal genomes from 2,445 breast cancers, sequenced by the 100,000 Genomes Project, sit in the National Genomic Research Library.
Inside the research environment, genomes are linked to clinical records (2,208 cases) and national mortality data (1,188 patients).
Pipelines profile each tumor for driver mutations, mutational signatures, homologous recombination deficiency, mismatch repair deficiency, and tumor mutational burden.
Survival models test which genomic features predict outcome; findings are checked against 1,803 patients from three independent cohorts, including Sweden’s SCAN-B.
What’s in the stack
Lifebit products used: Trusted Research Environment
03 · The results
From genome to treatment option
About 27% of tumors (656 of 2,445) had genetic features that could guide treatment now, with existing drugs or trial entry: more than 15,000 women a year in the UK. A further 15% had features useful for future drug development. In ER-positive, HER2-negative cancer, structural variation burden, high APOBEC signatures, and TP53 drivers each predicted survival independently of age, stage, and grade. The team built a risk tool from them and validated it in Sweden’s SCAN-B cohort. The authors propose a two-step model: check every genome first for targeted therapy or trial options, then use genome-based risk scores for everyone else.
Timeline
About this use case
Population genomics
Cancer genomics reads the full DNA of a patient’s tumor to find the changes that drive it and the drugs it may respond to. Here, whole genomes from NHS breast cancer patients were linked to their records and outcomes inside a secure environment. The output is a list of patients with treatable genetic features and a model that predicts who needs more, or less, treatment.
Who uses it here: A University of Cambridge team led by Serena Nik-Zainal, working with Genomics England, NHS colleagues, and three Lifebit co-authors, on patients recruited at 13 NHS Genomic Medicine Centres.
See how population genomics worksMore customer stories
100,000
participants in PRECISE-SG100K, Singapore’s national genome study (Chinese, Malay, and Indian)
Read the story →
5
funded NIH intramural research projects in 2026
Read the story →
59 min
for some target identification analyses that took months
Read the story →