Precise Diagnostics Boost for Cancer Patients

Qnomx is dedicated to advancing cancer care through precision genomic analysis. With years of experience interpreting complex cancer genomics cases worldwide, the company provides timely, expert analysis and consultation to healthcare providers, ensuring that world-class expertise is available when needed.

The Challenge

Qnomx, a startup specializing in cancer genomics interpretation, was grappling with an urgent bottleneck: lab reports from genomic profiling tests were 40–60 pages long and too complex for oncologists to process quickly. Even seasoned professionals struggled to interpret them accurately within tight clinical timelines, consuming 2–3 hours per report.

These delays risked incorrect or late treatment decisions, outcomes that cancer patients can’t afford. To improve this process, the goal was to develop an AI workflow that integrates retrieval (searching for relevant information) with generation (creating text based on retrieved information) to improve the quality of output and produce more informed and accurate results.

The Approach

For this case, several state-of-the-art open source components and a customized large language model (LLM) with local data storage and processing were used to improve the accuracy, explainability, and transparency of the data flow. An open source LLM was fine-tuned through supervised and reinforcement learning based on input from a genomics expert. Qnomx experts were able to provide the AI with the correct answers to typical genomic questions, which served as ground truth.

The model was then tested in multiple iterations on new datasets to evaluate its ability to perform the task. Through a Retrieval-Augmented-Generation (RAG), the user is able to upload a document (the cancer report) and chat with the document.

“The AI efficiently scans long cancer genomics reports (40- 60 pages) to help analysts capture important details and flag missing clinically relevant information for follow-up. It’s a great example of explainable AI in routine use today.” - James Creeden, Qnomx

The Result

The new solution resulted in a 90% reduction in text reading time and analysis work and a 24% increase in hardware efficiency. This hardware optimization was achieved by quantizing the model, making it leaner and more efficient. Finally, the AI was made highly trustworthy by significantly reducing hallucinations. In order to improve this even further, the model provides all the sources for the given answer in a visually appealing way within the uploaded report and has a self-check function. This allows the user to review only the relevant text in the report as needed.

Qnomx’s AI integration delivered remarkable real-world benefits:

• 90% decrease in time spent reading and analyzing reports.
• 8x improvement in lab throughput capacity.
• 325% increase in clinician satisfaction and confidence scores.

Next Steps

The investigation of how Quantum Reinforcement Learning and other Quantum machine learning techniques could enhance the computational pipeline for interpreting genomic reports will be explored in the next steps. Explanable Generative AI-based Cancer Genomics Report Interpretation reduces 90% of the work involved in reading and analyzing text.

For more insights into our collaboration, download our use case one pager or get in touch with our team.