Table 5 Diagnostic and predictive models built using genetic variables.
From: Clinical use of artificial intelligence in endometriosis: a scoping review
AI methods used | Authors [ref.] | Stage of endometriosis | Type of endometriosis | Sample size | Inputs used | Method accuracy |
---|---|---|---|---|---|---|
Deep Machine Learning Algorithm | Li et al.52 | All four stages of endometriosisa | Not specified | 213 patients (142 endometriosis, 71 controls) | SCAF11, KIF3A, KRAS, MDM2 | SE = 100% SP = 61.1% |
GenomeForest | Akter et al.53 | All four stages of endometriosisa | Not specified | Transcriptomics dataset: 16 endometriosis, 22 controls; methylomics dataset: 44 endometriosis, 36 controls | Genes in transcriptomics data and genomic regions in methylated data. 11 687 protein-coding genes (14 154 genes total) | For transcriptomics data: SE = 93.8% SP = 100% For methylomics data: SE = 92.9% SP = 88.6% |
Random-Forest-based Machine Learning Classification Analysis | Perrotta et al.54 | All four stages of endometriosisa | Not specified | 59 patients (35 endometriosis, 24 controls) | Operational taxonomic unit and community state types in vaginal microbiome | SE = NR SP = NR |
Decision Tree | Akter et al.55 | All four stages of endometriosisa | Not specified | Transcriptomics dataset: 38 samples (16 endometriosis, 22 controls); methylomics dataset: 77 samples (42 endometriosis, 35 controls) | Transcriptomics: 14 154 genes; methylomics: 2 577 382 methylated regions | For transcriptomics: SE = 81.3% SP = 95.5% For methylomics: SE = 76.2% SP = 80% |
Partial Least Squares Discrimination Analysis | Akter et al.55 | All four stages of endometriosisa | Not specified | Transcriptomics dataset: 38 samples (16 endometriosis, 22 controls); methylomics dataset: 77 samples (42 endometriosis, 35 controls) | Transcriptomics: 14 154 genes; methylomics: 2 577 382 methylated regions | For transcriptomics: SE = 86.4% SP = 56.3% For methylomics: SE = 60% SP = 76.2% |
Support Vector Machines | Akter et al.55 | All four stages of endometriosisa | Not specified | Transcriptomics dataset: 38 samples (16 endometriosis, 22 controls); methylomics dataset: 77 samples (42 endometriosis, 35 controls) | Transcriptomics: 14 154 genes; methylomics: 2 577 382 methylated regions | For transcriptomics: SE = 63.6% SP = 43.8% For methylomics: SE = 40% SP = 61.9% |
Random Forest | Akter et al.55 | All four stages of endometriosisa | Not specified | Transcriptomics dataset: 38 samples (16 endometriosis, 22 controls); methylomics dataset: 77 samples (42 endometriosis, 35 controls) | Transcriptomics: 14 154 genes; methylomics: 2 577 382 methylated regions | For transcriptomics: SE = 45.5% SP = 43.8% For methylomics: SE = 31.4% SP = 52.4% |
Margin Tree Classification | Tamaresis et al.56 | All four stages of endometriosisa | Not specified | 148 endometrial samples (77 endometriosis, 37 without endometriosis but other uterine/pelvic pathology, 34 controls) | FOSB, FOS, EGR1, JUNB, MTSS1L, CTSW, TGFB1, SOC3, IL32, FKBP8, ISYNA1, CCL3, GNLY, MAP3K11, C1QA, NOTCH3, CYR61, NPTXR, FBN1, PNRC2, ITGA6, DHFR, SLC39A6, MYO10, HSP90B1, SMC3, PKP4, PALLD, DIO2 | SE = NR SP = NR |