Imagine you have an intelligent friend who knows a lot about different topics. Ask them questions and you’ll get detailed and helpful answers. What is a large language model Similar to ChatGPT, but it’s a computer program instead of a human. Scientists train these large-scale language models on vast amounts of text and other information stored in websites, articles, and databases.
When we visit a doctor for an illness, we are often prescribed a combination of medications to treat our symptoms. Similarly, doctors try to determine which drug combinations will most effectively treat cancer patients. A combination of two or more drugs that is more effective than a single drug. Synergy of drug pairs.
researcher They used artificial intelligence (AI) techniques to predict the synergistic effects of drug pairs in common cancer tissues such as breast and lung cancer. However, limited experimental data stored in online databases and challenges in organizing data from published papers have led to the use of AI techniques to investigate rare cancers such as bone cancer. They couldn’t.
A group of researchers in the US sought to address the lack of AI research on rare cancers using large-scale language models. The researchers used drug pair synergy data from seven rare cancer tissues, including pancreas, endometrium, liver, soft tissue, stomach, urinary tract, and bone. drag comb database. Their drug pair synergy data consists of the name of the drug used to treat these cancers, the type of cancer used to treat it, and information about the synergy of the drug pair, less than or equal to 5. A higher score was given. A score of less than 5 means that the combination of two drugs does not work well in treating cancer, and a score of more than 5 means that the combination of two drugs is more effective. Masu.
First, they trained an AI to understand whether two drugs worked well together based on the drug pair’s synergy score. They prompted the AI ​​with phrases like: “Determine the synergistic effect of the following drugs in combination with anticancer drugs.” Then experiment with additional prompts that generate one-line answers, and Determine in one word whether the synergistic effect of the combination of drugs is positive.
After this prompt, They used 80% of the drug pair synergy data they had collected. from rare cancer Common cancers such as lung cancer, skin cancer, ovarian cancer, kidney cancer, breast cancer, as input. For example, to determine how effective her two drugs, AZD1775 and azacitidine, are in treating bone cancer, The first drug is AZD1775. The second drug is azacytidine. The tissue is bone. Synergy <5 inches. If the drug pair synergy score for these two drugs was greater than 5, they trained the model to output the word “positive.” This means that the combination of the two drugs effectively treats bone cancer. We trained the model to output the word “negative” if the drug pair synergy score for these two drugs was less than 5, meaning that the combination of the two drugs was not effective.
After training the AI, They used the same prompts and the remaining 20%. Using drug pair synergy data from common and rare cancers as input We test whether their model can predict positive and negative synergistic effects of drug pairs. They named their predictive AI model CancerGPT.
To evaluate CancerGPT’s performance, the researchers further investigated how accurately CancerGPT predicts positive and negative synergy of drug pairs on a 20% rare cancer tissue dataset. They are, Prediction of drug synergy CancerGPT had an accuracy of 90% for liver cancer, 86% for soft tissue cancer, and 81% for endometrial cancer. They also reported that the model predicted with less than 70% accuracy in datasets for urinary tract cancer, gastric cancer, and bone cancer.
The scientists concluded that CancerGPT is the first AI to predict the synergy of drug pairs in rare cancer tissues. They suggested that this new tool could help biologists develop drug combinations to treat rare cancer types. It also helps researchers design and perform synergy experiments for new drug pairs more quickly and cost-effectively. The researchers proposed further testing the model with drug pair synergy data from other infectious diseases to understand the benefits and accuracy of large-scale language models in treatment design.
Post views: 136
Source: sciworthy.com
