Applications of Machine Learning to Predict Cisplatin Resistance in Lung Cancer

Area of Science:

• Oncology
• Genomics
• Chemotherapy

Background:

• Lung cancer has the highest incidence and mortality globally.
• Platinum-based chemotherapy is crucial for lung cancer treatment, but patient response varies.
• Identifying patients resistant or sensitive to platinum-based therapy is essential for effective treatment.

Purpose of the Study:

• To identify a gene panel that predicts lung cancer cell line sensitivity to cisplatin.
• To explore potential mechanisms underlying cisplatin resistance or sensitivity.

Main Methods:

• Utilized drug response and sequencing data from 170 lung cancer cell lines (GDSC database).
• Employed support vector machines (SVMs) and beam search to select an optimal gene panel.
• Analyzed cell line data to investigate underlying biological mechanisms.

Main Results:

• A nine-gene panel (PLXNC1, KIAA0649, SPTBN4, SLC14A2, F13A1, COL5A1, SCN2A, PLEC, ALMS1) was identified with an AUC of 0.873 ± 0.004.
• Significantly higher lnIC50 values were observed in the mutant-type group compared to the wild-type group, irrespective of lung cancer subtype.
• Differentially expressed pathways between groups may explain the observed differences in sensitivity.

Conclusions:

• A nine-gene panel can accurately predict cisplatin sensitivity in lung cancer.
• This predictive panel holds potential for guiding individualized treatment strategies.
• Improved patient prognosis through personalized lung cancer therapy is a key implication.