The questions behind
precision medicine decisions
Every precision medicine programme reaches moments where the evidence is incomplete, the timelines are tight, and the next decision matters more than any that came before it.
This page explores the decisions translational teams face as programmes move from first patient data to the next development decision - and why answering them requires confidence in the evidence.
Decisions on this page
- ↓Can you explain who responded, and why?
- ↓Is your drug acting through the mechanism you expect?
- ↓How confident are you in a patient selection strategy built on one cohort?
- ↓Which biomarker would you take into the next phase?
- ↓Which patients belong in your next clinical trial?
- ↓Is your evidence traceable enough to defend?
- ↓What happens when the evidence arrives too late?
Every programme is different. These are some of the most common questions we explore with translational teams as evidence evolves.
Can you explain who responded, and why?
Yes, if molecular, imaging and clinical data are analysed together, not separately. Sonrai scientists bring the datasets into one analysis, compare responders and non-responders, and report which signals are associated with response and what the analysis cannot establish.
The clinical team has presented encouraging response data.
Leadership wants to understand why those patients responded.
The molecular data, pathology data and clinical outcomes all exist.
Different teams have analysed different parts of the programme.
Everyone has evidence.
No one has the whole picture.
Illustrative scenario, not a client account.
The tension
The clinical team is seeing response signals. The translational team still cannot confidently explain the biology. Molecular, imaging and clinical evidence exist - but they don't yet tell a coherent story.
Further reading
The real challenge isn't identifying responders. It's building enough integrated evidence to explain why they responded - with enough confidence to act on it.
Trying to understand who responded, and why?
Discuss treatment responseIs your drug acting through the mechanism you expect?
Target engagement shows a drug is biologically active. It does not show that this activity explains patient response. Sonrai scientists connect preclinical and clinical data to see whether the expected biological changes appear in responders, and report where the evidence supports the mechanism and where it does not.
Target engagement looks clear. The drug is doing something.
The preclinical data pointed strongly to a mechanism.
But as the clinical data comes in, the picture is less certain.
The expected biological changes are not consistently appearing across patients.
Or they are appearing - but not translating into clinical benefit the way the hypothesis predicted.
The team knows the drug is active. They are less sure it is active in the way that matters.
Illustrative scenario, not a client account.
The tension
Target engagement and clinical benefit are not the same thing. Many programmes can demonstrate that a drug is biologically active. Fewer can demonstrate that the biological activity they are seeing is the mechanism driving patient response.
Further reading
Mechanism of action evidence needs to be built continuously across the development lifecycle - not assumed from preclinical data and revisited only when clinical results are unexpected.
Trying to connect biological activity to patient outcomes?
Discuss mechanism of actionHow confident are you in a patient selection strategy built on one cohort?
A single well-designed dataset can often confirm or reject a hypothesis, for example a treated versus control comparison with enough patients and replicates. A patient selection strategy asks more, because it has to hold in patients the first cohort did not include. Sonrai scientists assess whether a candidate signal holds in additional cohorts or datasets where they are available, and report where it holds, where it weakens and why.
The initial analysis looked strong. The conclusion looked strong.
The team built a patient selection strategy around it.
Confidence grew. Timelines moved forward.
Then new data arrived - and the evidence didn't hold.
Now the programme is paused while the team tries to understand what changed.
Was it the biology? The patient population? The analytical approach?
No one can say with confidence.
Illustrative scenario, not a client account.
The tension
Many programmes look strong in early analysis and struggle when new evidence arrives. A next cohort, a different modality, clinical outcomes layered in - each can shift a conclusion that felt settled. A single well-designed dataset can answer a focused question. A patient selection strategy has to hold in patients the first cohort did not include, so the question is whether the signal would survive a different cohort.
Scientific perspective
Research studying transcriptomic biomarker signatures across multiple independent cohorts has found that predictive signatures for immunotherapy response rarely generalise beyond their discovery cohort - a challenge for any patient selection strategy built on a single cohort.
Amelia, Piyawajanusorn and Ballester - Briefings in Bioinformatics, 2025
Further reading
A selection strategy that does not hold in a new cohort was a hypothesis, not a strategy. Cross-cohort testing needs to be in place before the strategy depends on it.
Stress-testing your evidence base?
Discuss evidence testingWhich biomarker would you take into the next phase?
Compare the candidates on the same integrated evidence, not on whichever analysis was presented last. Sonrai scientists bring the datasets into one view, rank candidate biomarkers by the strength of the evidence, and set out the limits and recommended next analyses for each.
Different analyses support different biomarker candidates.
Different teams have reached different conclusions.
Each candidate has evidence behind it.
Each candidate has a question mark from a different direction.
The next board meeting is approaching.
Someone needs to decide which biomarker the programme advances on.
That decision will shape patient selection, trial design and the investment case for years.
Illustrative scenario, not a client account.
The tension
Biomarker prioritisation isn't just an analytical question - it's a programme commitment. Advancing on the wrong biomarker costs time, capital and credibility. But waiting for perfect certainty is also a decision, with its own costs. The question is whether the evidence base is strong enough to make the call.
Further reading
The biomarker you advance into the next phase should be the one with the strongest integrated evidence base - not the one that looked strongest in the most recent analysis.
Deciding which biomarker to advance?
Discuss biomarker prioritisationWhich patients belong in your next clinical trial?
Choosing the right patients is one of the hardest decisions in a trial, and no one has fully solved it. Start with the biology, then check what can be measured. Sonrai scientists compare candidate responder subgroups across molecular and clinical features, test whether they hold across datasets where available, and report which selection criteria the evidence supports, how strong that support is and what remains uncertain before it is used to select patients.
The first study is complete. The data is encouraging.
Now the team needs to define the population for the next trial.
The biology points to one group.
The tools available today can identify a different group.
They are not the same population.
Selecting on convenience is faster and cheaper.
But if the biology is right and the selection is wrong, the next trial will tell you - at significant cost.
Illustrative scenario, not a client account.
The tension
Biology versus convenience. The right patient population is defined by mechanism - by what the science says about who is most likely to respond. The convenient population is defined by what can be measured today with the tools and data already available.
Scientific perspective
Biomarker-driven enrichment designs require strong prior biological evidence to justify restricting enrolment to a defined subpopulation. Advancing patient selection criteria before that evidence base is established risks both enrolling the wrong population and losing the ability to evaluate the biomarker's true predictive value.
Dignam et al. - JCO Precision Oncology, 2019
Further reading
Patient selection built on convenience rather than biology doesn't just risk the next trial - it carries that risk silently forward until the data makes it visible.
Refining patient selection for your next trial?
Discuss patient stratificationIs your evidence traceable enough to defend?
Evidence is traceable when each finding can be followed back to the data version, method and workflow that produced it. Sonrai analyses run in versioned workflows inside a governed environment, with provenance recorded from raw data to finding.
Illustrative scenario 1
Board and investor readout
The slides look strong. The narrative is clear. But the board has a statistician on it. And the lead investor asked a question last quarter that no one answered cleanly. The analysis was done correctly. But if someone asked how it was done - who ran it, on which version of the data, using which method - the answer would take three days to reconstruct. The data is good enough to present. Whether it's good enough to defend is a different question.
Illustrative scenario 2
Regulatory or partner scrutiny
The programme is progressing. A regulatory submission is being prepared. Or a potential partner has asked for a data package before the next conversation. The analysis looks clean. The conclusions feel solid. But the methods were developed iteratively, across different team members, in different environments. There is no single document that describes exactly what was done, in what order, on which version of the data. The evidence is real. Whether it is traceable enough to submit or share is a different question.
The tension
Presenting evidence and defending a decision are different things. Presentation requires a compelling narrative. Defence requires knowing exactly how every finding was generated - and being able to reproduce it on demand. Most teams have the former. Fewer have the latter.
Further reading
The question isn't whether the science is right. It's whether you can show how it was produced - to an investor, a partner, or a reviewer who wasn't in the room when the analysis was run.
Preparing evidence for a critical programme decision?
Discuss translational evidenceWhat happens when the evidence arrives too late?
Slow answers usually come from data held in separate systems and analysis done step by step. Sonrai scientists work in one governed environment where datasets can be brought together, so a new question builds on existing work instead of starting again.
The clinical team raised a question about a subgroup of responders on Tuesday morning.
To answer it, three teams need to be involved.
Data needs to be pulled from four separate systems.
A CRO request has to be submitted and tracked.
The answer arrives three weeks later.
The decision it was meant to inform has already been made.
The data existed. The answer was always there. The infrastructure just couldn't deliver it in time.
Illustrative scenario, not a client account.
The tension
The gap between question and answer is not a competence gap - it is a capacity and access gap. The bottleneck is how long it takes to bring the data together and analyse it, not the quality of the people interpreting it.
Further reading
Slow cycle time isn't a sign of scientific rigour. It's a sign that there was no capacity to bring the evidence together when the question arose.
Trying to answer complex biological questions faster?
Discuss your workflowStart a conversation
Which of these questions is live in your programme?
Every programme is different. These are some of the most common conversations we have with translational teams. Tell us where you are and we'll follow up with relevant experience - not a sales pitch.
Thank you. Our team will be in touch with relevant experience from similar programmes, not a sales pitch.
Prefer to speak directly? Talk to our team.
These decisions require evidence that can stand up to scrutiny.
Not just data. Evidence that is integrated, traceable and built across the lifetime of a programme - so that when a decision needs to be made, the biological story holds together.
That's why our scientists work alongside precision medicine teams from translational research into clinical development. Where we develop custom code for your analysis, it is generally provided alongside the results.
Supported by a secure, reproducible analytical environment built specifically for precision medicine.