Machine learning on alignment features for parent-of-origin classification of simulated hybrid RNA-seq

BMC Bioinformatics

Table 6 Classification Performance with Reference Transcripts of genus Equus

Equus RNA	A Bowtie2	B Bo_AS	C Bo_RF	D STAR	E St_AS	F St_RF	G Salmon	H bwa
Accuracy Sensitivity Specificity Precision F1-score MCC AUPRC AUROC Pos Pref Ties	73.0% 78.4% 67.6% 70.8% 74.4% 0.463 – – 55.4% –	76.7% 75.2% 78.3% 77.6% 76.4% 0.535 – – 48.4% 38.6%	81.3% 91.1% 71.6% 76.2% 83.0% 0.637 92.0% 91.4% 59.7% –	78.8% 78.4% 79.1% 79.1% 78.7% 0.576 – – 49.6% –	77.2% 75.4% 78.9% 78.2% 76.8% 0.544 – – 48.2% 38.6%	85.8% 80.1% 91.5% 90.4% 84.9% 0.720 93.9% 93.8% 44.3% –	69.1% 54.8% 83.4% 76.8% 63.9% 0.399 – – 35.7% –	68.6% 76.7% 60.4% 66.0% 70.9% 0.376 – – 58.2% –

By many measures including accuracy, F1, and MCC, the random forest performance surpassed that of the other methods tested. For directional statistics, donkey and horse were considered the negative and positive classes, respectively. (A, B, C) Using the Bowtie2 aligner, a parent was chosen by the aligner, or by comparing alignment scores, or by the random forest, respectively. (D, E, F) Similar to A, B, and C but using the STAR aligner, configured to avoid splicing. (G, H) Parent chosen by Salmon or bwa, respectively

ISSN: 1471-2105