Table 1 Summary of work done on disease classification in metagenomics data using ML/DL methods
From: Automatic disease prediction from human gut metagenomic data using boosting GraphSAGE
Reference | ML/DL approach(es) used | Diseases considered | Input features | Remarks/observations |
|---|---|---|---|---|
[10] | SVM, RF | T2D, CRC, cirrhosis, IBD, obesity | OTU abundance | RF with feature selection outperformed basic RF and SVM classifiers with best AUC of 74%, 88.1%, 94.6%, 89.3%, 65.6% with T2D, CRC, cirrhosis, IBD, and obesity datasets respectively. |
[14] | SVM, RF, XGB, DF, AutoNN | T2D, CRC, cirrhosis, IBD, obesity | OTU abundance, k-mer frequency | The proposed AutoNN model achieved the best accuracy of 66.3% using OTU feature on T2D dataset. |
[15] | SVM | IBS | OTU abundance | The software package called metaDP can be used for classifying other disease samples. |
[16] | SVM, RF | Crohn’s disease | k-mer frequency | The best F1-score of 76% was achieved by RF classifier with k-mer feature. |
[17] | SVM, MLP | IBD | OTU abundance gene group abundance | The proposed hybrid classifier achieved an AUC of 80% |
[18] | MLP, CNN | IBD | OTU abundance | The model achieved the best AUC of 89% with MLP by using data augmentation technique. |
[19] | CNN | T2D, IBD, cirrhosis, CRC, obesity | OTU abundance | The model achieved the best accuracy of 84.2% and 66.3% for IBD and obesity datasets respectively. |
[20] | CNN | CRC | OTU abundance | The model achieved the best AUC of 75.7% with CNN . |
[21] | CNN | T2D, cirrhosis. | OTU abundance | The ensemble CNN achieved 76.2% AUC on T2D dataset and 91.1% on cirrhosis dataset |
[9] | CNN | IBD | OTU abundance | The model ph-CNN achieved the best Matthews Correlation Coefficient (MCC) of 92% |
[8] | CNN | IBD, T2D, obesity, cirrhosis | OTU abundance, phylogenetic relationship | The model PopPhy-CNN achieved the best F1-score of 58.7% for Obesity dataset. |