Deep-learning Artificial Intelligence (AI) has reshaped cancer research as well as the development of personalized clinical care. Advances in high-performance computing and the development of novel and innovative deep-learning architectures have led to a paradigm shift – dramatically affecting all aspects of oncology research, including the detection and classification of cancer, the molecular characterization of tumors and their microenvironment, drug discovery (and in some cases and repurposing of old drugs or drugs used in different therapeutic areas), and the prediction of treatment outcomes.
Now a new generation of AI applications, designed to allow rapid, low-cost detection of clinically actionable genomic alterations directly from tumor biopsy slides, has been developed by engineers and medical researchers at the University of California San Diego, CA. In a paper, published in the July 31, 2024 edition of the Journal of Clinical Oncology the authors describe the development and use of a new AI protocol, called DeepHRD, for examining routine biopsies.[1]
This design and the underlying research were funded by the U.S. National Institutes of Health (NIH), a Curebound Targeted grant, the UC San Diego start-up funding, and the UC San Diego Sanford Stem Cell Institute.
“The new [low-costs detection] method is designed to save weeks and thousands of dollars from clinical oncology treatment workflows for breast and ovarian cancers,” explained senior author Ludmil Alexandrov, Ph.D., professor of bioengineering and professor of cellular and molecular medicine at UC San Diego, CA.
Step forward
Developing these deep-learning AI tools, designed to complement or replace the expensive and time-consuming genomic testing required to determine the best first-line cancer treatment specific for individual patients, represents a major step forward in eliminating the delays and health inequalities – and disparities – that have confounded the promise of precision medicine for cancer patients around the world.
“A cancer patient today can expect to wait crucial weeks after their initial tumor diagnosis for a standard genomic test, resulting in life-threatening delays in treatment,” noted Alexandrov.
“It is very concerning that high costs and time delays render lifesaving treatment protocols inaccessible for most patients, disproportionately impacting resource-constrained settings,” he added.
Across campus
At UC San Diego, this work in developing a new detection methodology represents a collaboration between multiple departments, including the Department of Cellular and Molecular Medicine in the UC San Diego School of Medicine, the Shu Chien-Gene Lay Department of Bioengineering at the UC San Diego Jacobs School of Engineering, Institute of Engineering in Medicine, Department of Medicine, and the UC San Diego Moores Cancer Center.
“It was the potential of [improving the promise of] precision oncology to tailor an individual patient’s treatment options that motivated the collaborators,” explained Erik Bergstrom, Ph.D., lead author of the study and a postdoctoral researcher in Alexandrov’s lab, which bridges bioengineering and medicine.
“Unfortunately, high costs, tissue requirements, and slow turnaround times have, until today, hindered the widespread use of precision oncology, leading to suboptimal — and potentially detrimental — treatment for cancer patients,” Bergstrom added.
“We wanted to see if we could develop a completely different approach to resolve this serious issue by designing AI to circumvent the need for genomic testing,” he further noted.
Minimum amount of patient information
Bergstrom said the collaborators focused on leveraging the minimum amount of patient information that is available early in the diagnostic process. He explained that virtually every cancer patient undergoes a tumor biopsy, an invasive procedure where abnormal tissue is removed for cytological, histological, and molecular analyses, which is then routinely processed and examined. Although the process was developed in the late 19th century, tumor biopsies are still the gold standard and backbone technique in early clinical oncology workflows today that confirm if a tumor is benign or malignant.
“Our AI, applied directly to a traditional tissue slide, allows accurate, instantaneous detection of cancer genomic biomarkers,” Bergstrom said. He explained that the team focused on AI identification of a specific biomarker for homologous recombination deficiency (HRD), a condition in which a cancerous cell loses a specific DNA damage repair mechanism. [2]**
Detecting HRD generally involves screening for defects in relevant genes, including BRCA1 and BRCA2.[3]
Bergstrom further pointed out that cancer cells with HRD are generally sensitive to platinum therapies and targeted inhibition of poly-ADP ribose polymerase (PARP), a key component of alternative backup DNA repair pathways. Hence, identifying patients with ovarian cancer * or breast cancer cancer with HRD allows the identification of patients likely to benefit from PARP inhibitor therapies.
“This AI approach saves the patient critical time,” Alexandrov noted.
“Oncologists can prescribe treatment immediately after initial tissue diagnosis. Remarkably, the AI test has a negligible failure rate, while current genomic tests have a failure rate of 20% to 30%, necessitating re-testing, or even invasive re-biopsy.”
The study’s co-senior author Scott Lippman, M.D., UC San Diego distinguished professor of medicine, Center for Engineering and Cancer, and Moores Cancer Center member, said the new technology will remove barriers of time and money to allow immediate, universal access and equality to actionable genomic biomarker detection — required for precision therapy — for people with advanced cancers.
Training AI
In their study, the engineer and research team trained the DeepHRD technology in predicting HRD from hematoxylin and eosin (H&E)–stained histopathological slides, using primary breast cancers (n = 1,008) and ovarian cancers (n = 459) from The Cancer Genome Atlas (TCGA) and compared the results with four standard HRD molecular tests using breast cancer (n = 349) and ovarian cancer (n = 141) from independent data sets, including platinum-treated clinical cohorts with RECIST progression-free survival (PFS), complete response (CR), and overall survival (OS) endpoints. [1]
The outcomes of the study showed that DeepHRD predicted HRD with an Area Under the Curve (AUC) of 0.81 (95% CI, 0.77 to 0.85), a performance confirmed in two independent primary breast cancer cohorts (AUC, 0.76 [95% CI, 0.71 to 0.82]).[1]
In their analysis, the study authors concluded that the results obtained with DeepHRD, when compared with molecular testing, classified approximately 1.8- to 3.1 more patients with HRD, demonstrating the capability of better overall survival (OS) in high-grade serous ovarian cancer and platinum-specific progression-free survival (PFS) in metastatic breast cancer.
Closing the disparities gap
The extraordinary aspect of this breakthrough AI application is that it will benefit highly informed, -resourced populations, and remarkably, will close the severe disparities gap in precision medicine, especially in resource-constrained, remote regions worldwide where testing is not yet extant.
“The era of precision oncology took off in the late 1990s, but recent U.S. studies show that the vast majority of cancer patients are not receiving a precision therapy approved by the Food and Drug Administration (FDA),” Lippman said.
“And the prime reason for this is because they’re not getting tested. As a clinical oncologist — and I’ve been doing this for nearly 40 years — there is no question that this new approach is the future of precision oncology,” Lipman concluded.
The authors of the study expect that the same Deep-learning Artificial Intelligence technology behind DeepHRD, which is protected by provisional patents through UC San Diego, CA, and has been licensed to io9, an AI-driven digital pathology company, can be applied to most other genomic biomarkers and many forms of cancer.
__
Note:* HRD is a prognostic in Tipple Negative Breast Cancer (TNBC) and Ovarian Cancer.
** Various cancer types exhibit defects within the homologous recombination repair (HRR) machinery of the cell. HRR is a conservative mechanism which predominantly act in the S and G2 phases of the cell cycle (which, together with the G1 phase is known as the interphase – when the cell grows and makes a copy of its own DNA), and restores the original DNA-sequence at a site where a double-strand DNA break occurs. Impairment or loss of function of HRR, known as homologous recombination deficiency (HRD), occurs across most types of cancer. When this occurs, it forces cells to use other DNA-repair mechanisms, including the non-homologous end joining (NHEJ) pathway which uses specific proteins that recognize, resect, polymerize and ligate the DNA ends, allowing it function on a wide range of DNA-end configurations. This, in turn, leads to repaired DNA junctions that in most instances contain mutations. However, this ‘repair’ mechanism is also more error prone to ‘errors’ than the HRR mechanism. Today, the commonly known causes of HRD are loss of function mutations identified in a number of genes, including BRCA1, BRCA2, RAD51C, RAD51D, and PALB2. A number of other genes, as well as promoter hypermethylation of BRCA1, a key element of gene expression regulation is also identified as a possible cause of HRD.
Reference
[1] Bergstrom EN, Abbasi A, Díaz-Gay M, Galland L, Ladoire S, Lippman SM, Alexandrov LB. Deep Learning Artificial Intelligence Predicts Homologous Recombination Deficiency and Platinum Response From Histologic Slides. J Clin Oncol. 2024 Jul 31:JCO2302641. doi: 10.1200/JCO.23.02641. Epub ahead of print. PMID: 39083703.
[2] Doig KD, Fellowes AP, Fox SB. Homologous Recombination Repair Deficiency: An Overview for Pathologists. Mod Pathol. 2023 Mar;36(3):100049. doi: 10.1016/j.modpat.2022.100049. Epub 2023 Jan 10. PMID: 36788098.
[3] McGrail DJ, Li Y, Smith RS, Feng B, Dai H, Hu L, Dennehey B, Awasthi S, Mendillo ML, Sood AK, Mills GB, Lin SY, Yi SS, Sahni N. Widespread BRCA1/2-independent homologous recombination defects are caused by alterations in RNA-binding proteins. Cell Rep Med. 2023 Nov 21;4(11):101255. doi: 10.1016/j.xcrm.2023.101255. Epub 2023 Oct 30. PMID: 37909041; PMCID: PMC10694618.
Featured image by Ali Shah Lakhani on Unsplash. Used with permission
DOI:10.14229/onco.2024.08.02.001
 has reshaped cancer research as well as the development of personalized clinical care. Advances in high-performance computing and the development of novel and innovative deep-learning architectures have led to a paradigm shift - dramatically affecting all aspects of oncology research, including the detection and classification of cancer, the molecular characterization of tumors, and their microenvironment, drug discovery (and in some cases and repurposing of old drugs or drugs used in different therapeutic areas), to the prediction of treatment outcomes.){kind=link}