Great to see Fred Chong and his colleague Samantha Riesenfeld making significant strides with their project, Quantum Biomarker Algorithms for Multimodal Cancer Data! His leadership is driving innovative approaches to enhance biomarker identification in cancer research. Key highlights include improved biological relevance and valuable insights from real-world datasets. Excited for what’s next!
Here are the updates on performer 'Infleqtion' on their project 'Quantum Biomarker Algorithms for Multimodal Cancer Data' The main goal is to leverage quantum computational resources to tackle a fundamental challenge in cancer research: developing effective methods for biomarker identification within multimodal cancer data.? ? In Phase 1, the team developed a hybrid quantum-classical algorithm using polynomial constrained binary optimization (PCBO) to identify accurate biomarkers across multimodal biological data—DNA, mRNA, and pathomics—by capturing complex, higher-order feature correlations.?Over the course of Phase 2, the team achieved major advances that significantly enhanced the algorithm’s effectiveness while reducing quantum resource requirements by several orders of magnitude: ·?????????They introduced a new entropy-based cost function that improves the biological relevance of selected features by capturing correlations across multiple levels—first, second, and third order. ·?????????They applied the algorithm to real-world multimodal cancer datasets in contexts beyond the scalability of classical algorithms, identifying both established and underexplored biological drivers that may point to novel insights. ·?????????They have?dramatically reduced the total number of required calls to the quantum computer by leveraging parameter transfer techniques that initialize large quantum circuits using the results of smaller, tractable simulations. This pushed the classical simulations up to 32 qubits and created a concrete pathway towards future 50+ qubit demonstrations on real quantum hardware. ? The succinct, high-quality feature sets identified by this approach are significantly more amenable to clinical deployment than those identified by "black box" approaches lacking interpretability or overparameterized models, which often fail to generalize well in healthcare settings where patient data are inherently scarce. Beyond practical implementation advantages, these focused biomarker signatures reveal key information about underlying biological mechanisms that can guide future research and therapeutic development. The team is led by PI?Fred Chong?and co-PIs?Samantha Riesenfeld,? Alexander Pearson?(University of Chicago),?Aram Harrow(Massachusetts Institute of Technology), and?Teague Tomesh?(Infleqtion). ? #WellcomeLeap #QuantumForBio #FeatureSelection #ParameterTransfer #Cancerbiomarker ? ?