Population inference based on mitochondrial DNA control region data by the nearest neighbors algorithm.

Area of Science:

• Genetics
• Bioinformatics
• Anthropology

Background:

• Mitochondrial genome DNA sequences are commonly used for population and geographic assignment.
• Current methods face challenges in achieving fine-resolution subpopulation accuracy due to shared genetic ancestry and cohabitation.
• Accurate population assignment is crucial for forensic and anthropological studies.

Purpose of the Study:

• To evaluate the accuracy of population and subpopulation assignment using mitochondrial genome sequences and the K-nearest neighbors (KNN) algorithm.
• To compare the performance of KNN and K-weighted-nearest neighbors (KWNN) algorithms for ethnic inference.
• To provide an accurate method for both continental and subpopulation assignment.

Main Methods:

• Utilized sequence data from 3070 mitochondrial genomes, including training sets for continental populations (Europeans, Africans, Asians) and Taiwanese subpopulations.
• Applied the K-nearest neighbors (KNN) and K-weighted-nearest neighbors (KWNN) algorithms, weighted by genetic distance, for classification.
• Employed a testing sample set of 50 individuals to validate the accuracy of the population assignment methods.

Main Results:

• The training sample set achieved 82-99% accuracy for continental population assignment (K=1-101).
• Subpopulation assignment accuracy ranged from 54-77% (K=1-5), with four Taiwanese subpopulations exceeding 60% accuracy (Ami, Atayal, Saisiyat, Tao).
• The testing sample set demonstrated 94-100% accuracy for continental population assignment with recommended K values (5, 10, 35).

Conclusions:

• The KNN and KWNN algorithms provide accurate methods for assigning individuals to continental populations and finer-resolution subpopulations.
• This approach enhances the reliability of population assignment in forensic and anthropological research.
• The study successfully validated a precise methodology for both broad and specific ethnic inferences using mitochondrial DNA.