A DNA-derived phage nose using machine learning and artificial neural processing for diagnosing lung cancer

Area of Science:

• Biotechnology
• Nanotechnology
• Medical Diagnostics

Background:

• Existing electronic nose technologies are laboratory-bound, limiting rapid, field-deployable respiratory disease diagnostics.
• There is a need for portable, non-invasive diagnostic tools for respiratory illnesses.

Purpose of the Study:

• To develop a portable, pretreatment-free diagnostic system for respiratory diseases using a DNA-derived phage nose (D2pNose).
• To assess the efficacy of D2pNose in diagnosing lung cancer from raw human breath samples.

Main Methods:

• Engineered bacteriophages with DNA sequences from mammalian olfactory receptors to create the D2pNose.
• Verified phage surface chemistry through correlation analysis of calculated and experimental reactivity.
• Collected breath samples from healthy subjects and lung cancer patients for D2pNose exposure.
• Utilized deep learning and neural pattern separation for data analysis and classification.

Main Results:

• D2pNose demonstrated comprehensive reactivity mirroring mammalian olfactory receptors.
• Achieved over 75% diagnostic success rate for lung cancer.
• Attained over 86% classification success rate for lung cancer using raw human breath.
• Experimental validation confirmed the manipulated surface chemistry of engineered phages.

Conclusions:

• D2pNose presents a viable, portable solution for respiratory disease diagnosis, including lung cancer.
• The system's ability to analyze raw human breath without pretreatment is a significant advancement.
• D2pNose shows potential for application in diagnosing other respiratory diseases.