Exploration of identifying individual tumor tissue based on probabilistic model

Area of Science:

• Genetics and Genomics
• Forensic Science
• Oncology

Background:

• Tumor genomic variations present challenges in identifying the source of tumor samples.
• Accurate source identification is critical for clinical pathology and forensic investigations.

Purpose of the Study:

• To develop a probabilistic model for calculating likelihood ratios (LR) to address tumor sample source identification.
• To evaluate the model's performance using short tandem repeat (STR) genotyping data.

Main Methods:

• Developed a probabilistic model treating tumor tissue as a mixture of normal and tumor cells.
• Incorporated STR variant incidence (φ) and normal cell percentage (M) as a priori parameters.
• Analyzed STR genotyping data from tumor and reference blood samples of 17 colorectal cancer patients.

Main Results:

• All tumor samples yielded Log 10(LR) values greater than 10^14, indicating strong evidence of origin.
• Non-contributor tests showed 99.9% of quartiles with Log 10(LR) values less than 0.
• LR values increased with decreasing variant incidence (φ) and increasing normal cell percentage (M).

Conclusions:

• The probabilistic model effectively handles tumor allele variability for source identification.
• The model provides a robust evaluation of evidence strength for tumor origin determination.
• This approach enhances accuracy in clinical practice and forensic identification of tumor samples.