Table 1:
Disease class size before and after PCA-based data compression. Composite disease clusters (latent factors) are retained for latent factors whose explained variance exceed the explained variance of equivalent noise-based latent factors. PCA was applied separately for each disease class.
| Disease Category | Number of Phecodes | Number of latent factors | Total Explained Variance of latent factors (%) |
|---|---|---|---|
| Circulatory System | 155 | 4 | 50.1 |
| Congenital Anomalies | 54 | 1 | 38.3 |
| Dermatological | 94 | 4 | 45.3 |
| Digestive | 155 | 6 | 48.3 |
| Endocrine/Metabo lic | 146 | 2 | 51.5 |
| Genitourinary | 156 | 10 | 52.1 |
| Hematopoietic | 54 | 3 | 64.4 |
| Infectious Diseases | 55 | 2 | 44.7 |
| Injuries and Poisonings | 125 | 2 | 21.6 |
| Mental Disorders | 70 | 4 | 64.5 |
| Musculoskeletal | 120 | 2 | 36.0 |
| Neoplasms | 132 | 3 | 39.9 |
| Neurological | 78 | 5 | 53.5 |
| Pregnancy Complications | 44 | 1 | 22.9 |
| Respiratory | 75 | 6 | 64.0 |
| Sense Organs | 113 | 1 | 23.0 |
| Symptoms | 36 | 2 | 46.0 |
| Total | 1,662 | 58 | 45.1 (Mean) |