From the Editors: Artificial Intelligence in Cell and Gene Therapy

Recommend

by

Russell Y. Cruz, MD, PhD

Senior Editor, Telegraft

ChatGPT

OpenAI Language Model

In case you are among the few who haven’t spoken to ChatGPT yet, you’re in for a treat. Signing up here lets you check the features of this chatbot; see for yourself why people are excited. Since its launch, last year on November 30, 2022, this natural language Artificial Intelligence (AI) has grown its user base exponentially. Enough users converse with the chatbot that its servers are often at capacity.

ChatGPT is but one in the latest of encouraging news from the world of AI. Last year, AlphaFold has dramatically improved its predictive folding capabilities, further establishing itself as an indispensable tool, following its designation as Science’s Breakthrough of the Year in 2021.[i] With these advances, AI is becoming more entrenched in the scientific community. There are many advantages to closely integrating AI work in the cell and gene therapy space (with all its inherent complexity). Increasing our familiarity with it will ensure we get the most out of this platform.

Ask ChatGPT itself how AI can impact cell therapies and gene therapies, and you’ll get responses that sound like someone actually wrote them. The essay between divider asterisks gives one an idea. Besides combining the paragraphs across numerous chats, with the final paragraphs of some sessions removed to allow such a combination, no other edits were done in what ChatGPT wrote below:

***

Adoptive cell therapy (ACT) is a type of cancer treatment that involves removing immune cells from a patient's body, engineering them in some way, and then infusing them back into the patient's body. AI can be used to improve various aspects of ACT, including:

Cell selection: AI can analyze large amounts of data on cell behavior, allowing for the identification of specific cell subpopulations with desired properties. This can improve the selection of cells for ACT, increasing the chances of a successful treatment.
Cell engineering: AI can be used to optimize the genetic modification of cells, allowing for the creation of cells with improved therapeutic properties. For example, AI can be used to design and optimize CRISPR-based gene editing methods to create cells that are better able to target cancer cells.
Cell tracking: AI can be used to track the movement and behavior of cells in the body, which can help to understand how they interact with the patient's immune system and how they home in on tumors.
Dosage: AI can be used to optimize the dosage of cells used in ACT, which is important to ensure that the cells are effective while minimizing any potential side effects.
Combination therapy: AI can be used to identify potential combinations of ACT with other cancer treatments, such as chemotherapy or radiation therapy, to improve the overall effectiveness of treatment.

While AI has the potential to greatly improve cell and gene therapy research, there are also potential disadvantages to consider:

Bias: AI algorithms are only as good as the data they are trained on. If the data used to train the algorithm is biased, the algorithm will also be biased. This can lead to inaccurate or unfair results, particularly when it comes to patient selection.
Lack of interpretability: Some AI algorithms, such as deep learning neural networks, can be difficult to interpret. This can make it difficult to understand how the algorithm arrived at a particular decision or prediction, which can be a problem when it comes to understanding the mechanisms behind a therapy or identifying potential side effects.
Overreliance: There is a risk that AI will be used as a "black box" solution, where researchers may rely too heavily on the output of an AI algorithm without fully understanding the underlying science. This can lead to oversimplification of complex biological systems, and may not be able to fully capture the complexity of cell and gene therapy.
Data privacy: AI requires large amounts of data to train algorithms, and these data often include sensitive information such as patient medical records. Ensuring that this data is kept private and secure is an important consideration.
Costs: Implementing AI in cell and gene therapy research can be expensive, as it requires specialized hardware, software and expertise.

Yes, there are currently several clinical trials that are using AI in various ways. Some examples include:

AI-assisted diagnosis and treatment planning in radiology: AI algorithms are being used to analyze medical images, such as CT and MRI scans, to assist radiologists in diagnosing diseases and planning treatments.

AI-assisted drug discovery: AI is being used to analyze large amounts of data on genetic variations and drug interactions to identify new drug targets and predict the safety and efficacy of new drugs.

AI-assisted patient selection in CAR-T therapy: AI is being used to analyze patient data, such as medical history and genetic information, to identify the best candidates for CAR T cell therapy.

AI-assisted monitoring of cancer therapy: AI is being used to track the movement and behavior of T cells in the body and monitor the safety of CAR T cell therapy.

AI-assisted drug development: AI is being used to optimize the design of vectors for gene therapy, and to identify potential combinations of gene therapy with other therapies to improve the overall effectiveness of treatment.

These are a few examples of how AI is being used in clinical trials, and the number of clinical trials using AI is expected to increase in the future. It's important to note that AI is not intended to replace clinical judgement but rather to support it.

AI has the potential to automate certain tasks that are currently performed by data scientists, such as data cleaning, feature selection, and model selection. However, it's important to note that AI is not intended to replace data scientists, but rather to assist them in their work. Data scientists are still needed to design, implement, and interpret the results of AI models.

Data scientists have a unique set of skills that include statistical analysis, programming, and domain-specific knowledge, which are vital for the design and implementation of AI models. They are able to understand the underlying data and the business problem and can use their expertise to design the appropriate AI solution. They also understand the underlying assumptions of the AI models and can interpret the results, which is crucial for decision-making and for avoiding biases.

Moreover, AI models require continuous monitoring, maintenance, and updating as the data changes, which is also a task that data scientists can perform. Additionally, Data scientists will still be needed to identify new opportunities to apply AI, and to develop new methods and models that are more effective than existing ones.

In summary, while AI has the potential to automate certain tasks performed by data scientists, it is not intended to replace them. Data scientists are still needed to design, implement, and interpret the results of AI models, and to identify new opportunities to apply AI.

***

(This is me again, Russell, taking over from ChatGPT). While writing such as the example above has generated concerns from the scientific community, with concerns about the “flood of AI-generated papers,”[i] for example as a recent editorial in Nature points out, incorporating this technology will be more advantageous than disadvantageous. We just have to be aware of its limits.

References

[1] https://www.science.org/content/article/breakthrough-2021

[1] The AI writing on the wall. Nat Mach Intell 5, 1 (2023). https://doi.org/10.1038/s42256-023-00613-9