he reviewer of cancer research history can only marvel at the acceleration of drug discovery and development plus the multimodal and multidisciplinary innovations. The landscape imagines incredible potential, while maintaining ethical guidelines. Here, cancer treatment advances are explored with the future of artificial intelligence (AI) and machine learning (ML).

20th Century

In the early 1900s, Paul Ehrlich developed drugs to treat infectious diseases, coining the term “chemotherapy.” Modern chemotherapy is attributed to the introduction of sulfur mustard gas as a chemical weapon in World War I, leading to the discovery of its antiproliferative properties as well as those of other compounds that interfered with tumor cell proliferation.

First-generation anticancer drugs discovered in the mid-1900s changed the clinical oncology landscape with the transient benefits and rare cures leading to combinatorial chemotherapy as standard of care. Impressive clinical outcomes of combinations in childhood acute lymphoblastic leukemia led to applications in other malignancies, including chemotherapy regimens like R-CHOP and FCR for blood cancers and carboplatin-Taxol or gemcitabine-cisplatin for solid tumors.

The first biologics, monoclonal antibodies (mAbs), targeted tumor cells more effectively and with less toxicity than chemotherapy. The first approved mAb was rituximab for lymphoma treatment, with current mAbs seeing wide clinical use for treating colorectal and breast cancers, respectively, and also being employed in cancer therapy in combination with chemotherapy and as a component of more complicated therapeutics as described below. Bone marrow transplantation was the first immunotherapy and is a method for transferring a system for the creation of blood cells together with an immune system from a healthy donor into a patient to control leukemia, either following initial treatment or for relapsed disease. It remains the standard of care across various leukemia subtypes for the purpose of seeking a cure. Advances in pharmacology and less toxic preconditioning regimens have led to a series of stepwise improvements, both in increased graft acceptance and reducing graft-versus-host disease incidence and severity.

21st Century

At the dawn of the 21st century, targeted therapies and advanced immunotherapies have provided researchers with new tools to treat cancers. Although first conceived in the 1970s and 1980s, advances in molecular and cell biology in more recent years have resulted in new cancer therapies. Unlike chemotherapy, which kills both healthy and cancerous cells, targeted therapy exploits properties unique to cancer cells, like changes to the cell surface, DNA repair deficiency, or the machinery that induces apoptosis, in order to kill those cells selectively.

• Kinases are enzymes that make changes to cellular proteins; however, mutated versions can become constitutively active and result in cancer initiation and growth. Inhibitors impacting kinase function are valuable in treating various cancers. Multikinase therapies are typically small molecule inhibitors that target more than one kinase simultaneously. This strategy can be helpful in overcoming drug resistance but can also give rise to adverse events.
• Antibody drug conjugates (ADCs) consist of three parts: an antibody component specific for the targeted cancer type, a cytotoxic chemotherapeutic component that destroys the cancer cell, and a linker component that connects the antibody to the cytotoxic chemotherapeutic. A main advantage of ADC is that chemotherapeutics bind cancer cells preferentially over normal cells.
• An immunotherapy that uses mAbs as checkpoint inhibitors has become a reliable and effective therapeutic option for several tumor types. By effectively blocking these checkpoint proteins, the T cell identifies the cancer cell as a threat and destroys it. However, immune-related adverse events are a significant side effect.
• Adoptive T cell therapies encompass a group of treatments that utilize different cell types from the immune system to destroy cancer cells. They include tumor-infiltrating lymphocyte, engineered T cell receptor, and chimeric antigen receptor (CAR)-T cell therapies that bind to specific antigens on the surface of cancer cells.
• CAR-T therapies have achieved unprecedented responses with hematologic malignancies, such as relapsed/refractory B-cell acute lymphocytic leukemia (leukemia returning after remission), non-Hodgkin lymphoma, and multiple myeloma. Solid tumors are particularly challenging targets for CAR-T therapy. Patients undergoing CAR-T therapy must be monitored closely, as life-threatening adverse events may occur. Additionally, relapse can follow successful treatment with CAR-T therapies.
• Bispecific T cell engagers (BiTEs) have shown success in treating hematologic malignancies and are designed to redirect T cells to specific tumor antigens and activate the T cells directly and proximally to the cancer cells.

Table 1. Classes of FDA-Approved Cancer Drugs of the 21st Century.

The Future: AI/ML

With advances in bioinformatics, spatial histology, noninvasive bioanalytical analyses, electronic medical records, wearable monitors, and large computing power, the use of AI and LLM is necessary for progress in cancer care. Early detection and intervention are improving with advances in surgery, radiation, and vaccine development. For inoperable and disseminated disease, AI can be harnessed by multiplexing large data sets towards optimizing diagnostics, predictions, and treatment options to deploy our armamentarium of therapeutics.

Early detection requires reliable and sensitive methods to detect pre-cancerous cells, which optimally can be done as a point-of-care practice. A negative predictive test avoids expensive radiological and invasive biopsies. A positive predictive test can inform decisions about best treatment options. AI supports a continuously virtuous learning algorithm operating in multiple dimensions enabling scientists and clinicians eventually to cure cancers, one by one.

Recent work has begun to move the field of AI and early detection forward. Findings published in Nature Medicine last year suggest that a deep learning algorithm can predict the risk of pancreatic cancer up to three years before diagnosis. Similarly, the recently developed AI tool Sybil may detect future lung cancer risk using low-dose CT scan data. In the interception space, another deep learning model uses radiology text reports to estimate Response Evaluation Criteria in Solid Tumors (RECIST) scores and measure objective response categories for clinical trials. Others are using AI and machine learning to determine immune interactions and tumor heterogeneity in melanomas as well as colorectal, lung, breast, and neuroendocrine tumors.

As these and future models mature and evolve to include more cancer types and greater computational capabilities, they have the potential to revolutionize cancer treatment.