Research ArticlePRECISION MEDICINE

Identification of type 2 diabetes subgroups through topological analysis of patient similarity

See allHide authors and affiliations

Science Translational Medicine  28 Oct 2015:
Vol. 7, Issue 311, pp. 311ra174
DOI: 10.1126/scitranslmed.aaa9364
  • Fig. 1. Patient and genotype networks.

    (A) Patient-patient network for topology patterns on 11,210 Biobank patients. Each node represents a single or a group of patients with the significant similarity based on their clinical features. Edge connected with nodes indicates the nodes have shared patients. Red color represents the enrichment for patients with T2D diagnosis, and blue color represents the nonenrichment for patients with T2D diagnosis. (B) Patient-patient network for topology patterns on 2551 T2D patients. Each node represents a single or a group of patients with the significant similarity based on their clinical features. Edge connected with nodes indicates the nodes have shared patients. Red color represents the enrichment for patients with females, and blue color represents the enrichment for males.

  • Fig. 2. Genotype-phenotype network for three subtypes in T2D.

    The network consists of the significant association between phenotypes and genetic variants at gene level specific to three T2D subtypes (subtype 1 in blue, subtype 2 in orange, and subtype 3 in pink). Phenotypes (oval) and genes (triangle) are connected by gray lines (P value). Oval nodes in dark green indicate the shared phenotypes across subtypes. The edge width reflects the significance of the P value for enrichment. The size of the node reflects the amount of associated genes or phenotypes. This network was visualized using Cytoscape 3.2.0.

  • Table 1.

    Clinical variables specific to subtypes. S-1, subtype 1; S-2, subtype 2; S-3, subtype 3; BMI, body mass index.

    (A) Clinical variables significantly specific to T2D subtype 1
    Clinical variablesMean or % subtype 1Mean or % subtype 2Mean or % subtype 3P (1 versus 2 + 3)S-1S-2S-3
    Platelet count (109/liter)98.36 ± 17.86228.24 ± 2.90228.61 ± 2.45<0.0001Y
    Urine protein concentration (mg/dl)51.19 ± 14.38152.67 ± 37.21219.98 ± 47.620.0001Y
    Lactate dehydrogenase (U/liter)193.35 ± 8.88231.03 ± 8.82251.34 ± 8.17<0.0001Y
    Age (years)59.76 ± 0.4564.25 ± 0.5063.65 ± 0.38<0.0001Y
    Blood urea nitrogen (mg/dl)16.69 ± 0.3519.38 ± 0.5919.52 ± 0.35<0.0001Y
    Neutrophil count (109/liter)2.50 ± 0.584.78 ± 0.124.83 ± 0.090.0024Y
    White blood cell count (109/liter)5.32 ± 0.577.28 ± 0.097.46 ± 0.070.001Y
    Respirations16.65 ± 0.1617.50 ± 0.1417.62 ± 0.08<0.0001Y
    Urine protein-to-creatinine ratio0.40 ± 0.091.19 ± 0.262.48 ± 0.45<0.0001YY
    Serum creatinine (mg/dl)1.00 ± 0.021.25 ± 0.071.27 ± 0.04<0.0001Y
    Eosinophil count (109/liter)0.09 ± 0.020.19 ± 0.010.20 ± 0.010.0003Y
    Blood protein total (g/dl)7.49 ± 0.037.34 ± 0.047.14 ± 0.03<0.0001YY
    Serum albumin (g/dl)4.27 ± 0.024.03 ± 0.034.04 ± 0.02<0.0001Y
    Serum calcium (mg/dl)9.90 ± 0.029.66 ± 0.039.60 ± 0.02<0.0001Y
    CO2 total26.60 ± 0.1326.05 ± 0.1526.16 ± 0.090.0011Y
    Mean platelet volume (fl)9.97 ± 0.378.98 ± 0.058.97 ± 0.040.008Y
    Prothrombin time* (s)29.18 ± 3.6414.10 ± 0.3314.13 ± 0.270.0005Y
    INR*2.57 ± 0.341.19 ± 0.039.32 ± 0.410.0005Y
    BMI33.07 ± 0.2931.32 ± 0.3031.19 ± 0.02<0.0001Y
    Estimated GFR calculation (MDRD, ml/min/1.73 m2)74.86 ± 1.4768.40 ± 1.9965.04 ± 1.33<0.0001Y
    GFR estimate (ml/min/1.73 m2)72.26 ± 1.4764.62 ± 1.7763.75 ± 1.22<0.0001Y
    Glucose* (mg/dl)193.69 ± 11.45149.55 ± 4.18158.69 ± 2.900.0005Y
    Insulin21.92%29.82%45.16%<0.0001YYY
    Metformin6.43%23.01%21.17%<0.0001Y
    Loop diuretics5.51%14.10%18.34%<0.0001Y
    DPP41.05%6.48%6.39%<0.0001Y
    CCBs19.55%30.63%35.31%<0.0001Y
    β-Blocker21.92%39.06%45.80%<0.0001YY
    ARB/ACEI48.16%57.05%62.96%<0.0001YY
    Vasodilators0.92%5.02%5.57%<0.0001Y
    Nicotinic acid derivatives0.13%1.30%1.37%0.02Y
    (B) Clinical variables significantly specific to T2D subtype 2
    Clinical variableMean or % subtype 1Mean or % subtype 2Mean or % subtype 3P (2 versus 1 + 3)S-1S-2S-3
    Weight (kg)92.26 ± 1.0885.17 ± 1.1489.16 ± 0.83<0.0001Y
    Troponin I level (ng/ml)00.03 ± 0.010.36 ± 0.090.0003YY
    Insulin21.92%29.82%45.16%<0.0001YYY
    (C) Clinical variables significantly specific to T2D subtype 3
    Clinical variableMean or % subtype 1Mean or % subtype 2Mean or % subtype 3P (3 versus 1 + 2)S-1S-2S-3
    Blood protein total (g/dl)7.49 ± 0.037.34 ± 0.047.14 ± 0.030YY
    Urine protein-to-creatinine ratio0.40 ± 0.091.19 ± 0.262.48 ± 0.450.0006YY
    Troponin I level (ng/ml)00.03 ± 0.010.36 ± 0.090.0003YY
    Systolic blood pressure (mmHg)132.04 ± 0.73132.41 ± 0.92135.7 ± 0.70.0001Y
    Serum chloride level (mEq/liter)101.01 ± 0.17101.45 ± 0.18102.03 ± 0.110Y
    HMG-CoA reductase inhibitors (statins)42.26%45.71%56.39%<0.0001Y
    Centrally acting antihypertensives1.44%1.30%4.11%0.0001Y
    ARB/ACEI48.16%57.05%62.96%<0.0001YY
    β-Blocker21.92%39.06%45.80%<0.0001YY
    Insulin21.92%29.82%45.16%<0.0001YYY

    *Point of care.

    • Table 2.

      Significant associated disease categories. MHSA, mental health and substance abuse; LCI, lower confidence interval; UCI, upper confidence interval.

      (A) Significant disease categories associated with T2D subtype 1
      Disease categoryRR95% LCI95% UCIP value
      Other upper respiratory infections1.681.342.11<0.0001
      Immunizations and screening for infectious disease1.651.322.06<0.0001
      Diabetes mellitus with complications1.501.221.840.0001
      Other skin disorders1.411.131.760.003
      E codes: place of occurrence1.381.081.770.01
      Blindness and vision defects1.321.041.670.02
      Other screening for suspected conditions (not mental disorders or infectious diseases)1.281.041.580.02
      Screening and history of MHSA codes0.740.590.940.01
      Other circulatory disease0.680.540.870.002
      Acute and unspecified renal failure0.630.420.940.02
      Pulmonary heart disease0.600.370.980.04
      Deficiency and other anemia0.570.450.71<0.0001
      E codes: adverse effects of medical care0.550.380.790.001
      Coronary atherosclerosis and other heart disease0.510.400.64<.0001
      Peri-, endo-, and myocarditis; cardiomyopathy (without tuberculosis or sexually transmitted disease)0.480.280.820.01
      Aortic, peripheral, and visceral artery aneurysms0.360.210.640.0004
      HIV infection0.220.120.38<0.0001
      (B) Significant disease categories associated with T2D subtype 2
      Disease categoryRR95% LCI95% UCIP value
      Cancer of bronchus: lung3.761.1412.390.03
      Malignant neoplasm without specification of site3.461.239.700.02
      Tuberculosis2.931.306.640.01
      Coronary atherosclerosis and other heart disease1.281.011.610.04
      Other circulatory disease1.271.021.580.03
      Age1.011.001.020.003
      Allergic reactions0.700.570.850.0004
      Other screening for suspected conditions (not mental disorder or infectious disease)0.640.520.79<0.0001
      Disorders of lipid metabolism0.560.450.70<0.0001
      E codes: struck by; against0.410.180.920.03
      Peritonitis and intestinal abscess0.120.020.880.04
      (C) Significant disease categories associated with T2D subtype 3
      Disease categoryRR95% LCI95% UCIP value
      HIV infection1.921.302.850.001
      E codes: adverse effects of medical care1.841.412.39<0.0001
      Aortic and peripheral arterial embolism or thrombosis1.791.182.710.01
      Hypertension with complications and secondary hypertension1.661.292.15<0.0001
      Coronary atherosclerosis and other heart disease1.411.151.720.001
      Allergic reactions1.421.191.700.0001
      Deficiency and other anemia1.391.141.680.001
      Screening and history of MHSA codes1.301.071.580.01
      Diabetes mellitus with complications0.800.670.960.02
      E codes: place of occurrence0.710.560.890.003
      Other upper respiratory infections0.730.570.920.01
      Blindness and vision defects0.710.570.880.002
      Other skin disorders0.680.550.830.0003
    • Table 3. Significant phenotypes.
      (A) Significant phenotypes with disease–genetic variant enrichments
      specific to T2D subtype 1
      PhenotypesGene symbolP
      Albumin-to-creatinine ratiosACE1.00 × 10−27
      Aspartyl phenylalanine levelsACE1.00 × 10−27
      B cell countLAMB41.00 × 10−27
      Chronic heart failureLEPR1.00 × 10−27
      Crypt frequencySEMA3A1.00 × 10−27
      DyslexiaCLSTN21.00 × 10−27
      HypercholesterolemiaBTN2A11.00 × 10−27
      Mannose-binding lectin levelsMBL21.00 × 10−27
      Prominence of right endocanthionTMTC21.00 × 10−27
      Retinol levelsFFAR41.00 × 10−27
      Phosphorylated τ 181 protein levelsMTUS1, UNC5C5.53 × 10−3
      Angiotensin-converting enzyme activityACE1.32 × 10−2
      Diabetes mellitusBTN2A11.32 × 10−2
      Entorhinal cortical volumeF13A11.32 × 10−2
      Multiple system atrophySNCA1.32 × 10−2
      N-acetylornithine levelsALMS11.32 × 10−2
      OtosclerosisTGFB11.32 × 10−2
      Pelvic organ prolapseZFAT1.32 × 10−2
      Tanning abilityMC1R1.32 × 10−2
      Vitamin D concentrationsGC1.32 × 10−2
      Diabetic retinopathyPLXDC2, HS6ST32.32 × 10−2
      Alanine transaminase levelsZNF8273.66 × 10−2
      Diabetic nephropathyACE3.66 × 10−2
      Left ventricular wall thicknessGRID13.66 × 10−2
      Leptin receptorLEPR3.66 × 10−2
      Forced expiratory volumeZSCAN31, TNS15.00 × 10−2
      Platelet response to aspirin
      intervention therapy
      ZNF583, GLIS35.00 × 10−2
      (B) Significant phenotypes with disease–genetic variant enrichments specific to
      T2D subtype 2
      PhenotypesGene symbolP
      Alcohol and nicotine codependencePLEKHG11.00 × 10−27
      Bleomycin sensitivitySAMD121.00 × 10−27
      Epirubicin-induced adverse drug reactionsMCPH11.00 × 10−27
      Follicular lymphomaSV2B1.00 × 10−27
      Lactose intoleranceST51.00 × 10−27
      Pronasale to left alare distanceCACNA2D31.00 × 10−27
      Stem cell transplantationNLRP31.00 × 10−27
      Geographic atrophyHTRA1, CFH6.57 × 10−4
      BrainCDH47.58 × 10−3
      Left ventricular internal diastolic dimensionsSLC35F17.58 × 10−3
      Mean platelet volumeARHGEF37.58 × 10−3
      Polypoidal choroidal vasculopathyCFH7.58 × 10−3
      PsychosisZNF804A7.58 × 10−3
      Suicidal behaviorGFRA17.58 × 10−3
      Tanning abilityHERC27.58 × 10−3
      Total τ protein levelsCDH47.58 × 10−3
      Meningococcal diseaseTMPRSS15, CFHR3, CFH7.79 × 10−3
      KeratoconusSOX5, MACROD21.76 × 10−2
      MeningiomaCHN22.14 × 10−2
      Polycystic ovary syndromeDENND1A2.14 × 10−2
      Primary sclerosing cholangitisGAS72.14 × 10−2
      Atrial fibrillationCAV1, HCN42.64 × 10−2
      Age-related macular degenerationPLEKHA1, HTRA1, IL8, CFH3.09 × 10−2
      Open-angle glaucomaADAMTSL1, CAV13.71 × 10−2
      Phosphorylated τ 181 protein levelsCHN24.04 × 10−2
      (C) Significant phenotypes with disease–genetic variant enrichments specific to
      T2D subtype 3
      PhenotypesGene symbolP
      Gallbladder cancerCNTN4, DCC1.00 × 10−27
      AllergyFHIT1.00 × 10−27
      B cell chronic lymphocytic leukemiaCD381.00 × 10−27
      Lymphoid interstitial pneumonitisFGF141.00 × 10−27
      OsteoporosisALDH7A11.00 × 10−27
      Peripartum cardiomyopathyAKAP131.00 × 10−27
      RR intervalGPR1331.00 × 10−27
      Spinocerebellar ataxia type 1ATXN11.00 × 10−27
      Intraventricular septal thicknessEXT1, CERS61.65 × 10−3
      Endometrial cancerSLC8A11.40 × 10−2
      HIV-associated
      neurocognitive disorders
      SLC8A11.40 × 10−2
      Response to statinASB181.40 × 10−2
      Uterine leiomyomaTNRC6B1.40 × 10−2
      Vitamin D concentrationsDAB11.40 × 10−2
      Anxiety disordersSDK2, FHIT2.50 × 10−2
      Cognitive declineCTNND23.86 × 10−2
      DementiaABCA13.86 × 10−2
      Estrone levelsESR13.86 × 10−2
      Impaired play skillsDCC3.86 × 10−2
      IntelligenceCNTN43.86 × 10−2
      MyopiaMIPEP3.86 × 10−2
      Plasma progranulin levelsDNAH113.86 × 10−2
      Polycystic ovary syndromeTHADA3.86 × 10−2
      Renal cell carcinomaITPR23.86 × 10−2
      Theta power of electroencephalogramST6GALNAC33.86 × 10−2
      Central corneal thicknessCOL5A1, FNDC3B4.00 × 10−2
      Atrial fibrillationC9orf3, SYNE25.00 × 10−2
      DepressionFHIT, BICC15.00 × 10−2
    • Table 4.

      Canonical pathways at gene level for each T2D subtype. ns, not significant.

      Canonical pathwaySubtype 1Subtype 2Subtype 3Genes
      Fatty acid β-oxidation III1.1 × 10−3nsnsECI1, ECI2
      Acetate conversion to acetyl-CoA3.5 × 10−3nsnsACSL1, ACSL2
      Netrin signaling6.2 × 10−3nsnsABLIM1, PRKG1, UNC5B, UNC5C
      GABA receptor signaling8.8 × 10−3nsnsADCY8, ALDH5A1, GABBR1, GABRR2, GPHN
      cAMP-mediated signaling9.2 × 10−3ns2.0 × 10−2Subtype 1: ADCY8, AKAP12, CAMK1D, CNGB1,
      CNGB3, GABBR1, MC1R, PDE3A, PKIA, RGS7
      Subtype 3: AKAP13, CAMK4, CHRM5, GNAI3, HTR1D,
      PDE4B, PDE6A, PRKAR2B, RAF1
      Role of pattern recognition receptors in recognition
      of bacteria and viruses
      ns1.8 × 10−3nsCXCL8, MAPK10, NLRP3, OAS1, OAS3, PRKCD, PRKCH
      Thrombopoietin signalingns6.8 × 10−3nsGAB2, PRKCD, PRKCH, SOS1
      α-Adrenergic signalingnsns1.2 × 10−3CAMK4, GNAI3, GYS1, ITPR2, PRKAR2B, RAF1, SLC8A1
      Synaptic long-term depressionnsns1.4 × 10−3GNA11, GNAI3, GRID2, GRM1, ITPR2, PLA2G4C,
      PLA2R1, PPP2R5B, RAF1,
      CREB signaling in neuronsnsns1.4 × 10−3CAMK4, GNA11, GNAI3, GRID2, GRIK4, GRM1, ITPR2,
      POLR2I, PRKAR2B, RAF1
      Glutamate receptor signalingnsns4.2 × 10−3CAMK4, GRID2, GRIK4, GRM1, PICK1
      Hepatic fibrosis/hepatic stellate cell activation3.0 × 10−2ns4.0 × 10−3Subtype 1: BCL2, COL19A1, COL28A1, IGF1R, IL1RAP,
      LEPR, TGFB1, TGFB2
      Subtype 3: BAX, COL15A1, COL25A1, COL4A4, COL5A1,
      COL5A3,COL9A3, FGF2, KLF12, MYH7B
      Sperm motilitynsns7.3 × 10−3CAMK4, ITPR2, PDE4B, PLA2G4C, PLA2R1, PRKAR2B, SLC12A2
    • Table 5. Toxicity functions at the gene level for each T2D subtype.
      Toxicity functionsSubtype1Subtype2Subtype3Genes
      Biliary hyperplasia3.5 × 10−3nsnsCFTR, PKHD1
      Glutathione depletion in liver3.5 × 10−3nsnsLEPR, TGFB1
      Liver fibrosis3.5 × 10−3nsnsTGFB1, LEPR, TGFB2, PKHD1
      Glomerular injury4.7 × 10−3nsnsFYN, TGFB1, LEPR, RARA, TNS1, PKN1, PTGER1, BCL2
      Renal hypertrophy4.7 × 10−3nsnsTGFB1, LEPR, RARA, BCL2
      Liver damage5.1 × 10−3nsnsSLC10A1, TGFB1, IGF1R, GABBR1, SERPINA1,
      CD274, PARK2, PTGER1
      Liver inflammation/hepatitis5.1 × 10−3nsnsAKAP12, SLC10A1, TGFB1, PDE3A, IGF1R,
      GABBR1, CD274, PARK2
      Renal proliferation7.6 × 10−3nsnsPRKG1, TGFB1, UNC5B, TTLL4, CRK, ZNF512B,
      DLC1, BCL2, UNC5C, AFF1
      Renal degeneration8.0 × 10−3nsnsTGFB1, TNS1, BCL2
      Cardiac arrhythmians1.0 × 10−3nsKCND3, HCN4, KCNG2, KCNQ1, CNTN5
      Bradycardians4.9 × 10−3nsHCN4, KCNQ1
      Cardiac arteriopathyns9.3 × 10−34.8 × 10−6Subtype 2: SAMD12, KALRN, ITGA8, PDE5A, DOCK4,
      CNTN6, PRKCH, CSMD2, CPEB3, CNTN5
      Subtype 3: CERS6, CLIC5, ZMYM2, CDCP1, ABCG1, FRMD4A,
      PDE4B, PTPRM, ABCA1, F2, SPATA5, AKAP13, MCF2L, PBX3, CNTNAP5,
      FMN2, CACNA2D1, SLC8A1, ESR2
      Liver fibrosisnsns3.3 × 10−3FGF2, PLAUR, BMP7, CC2D2A, F2, HSPB1
      Congenital heart anomalynsns5.8 × 10−3DNAH11, BICC1, PDS5B, INVS

    Supplementary Materials

    • Supplementary Material for:

      Identification of type 2 diabetes subgroups through topological analysis of patient similarity

      Li Li, Wei-Yi Cheng, Benjamin S. Glicksberg, Omri Gottesman, Ronald Tamler, Rong Chen, Erwin P. Bottinger, Joel T. Dudley*

      *Corresponding author. E-mail: joel.dudley{at}mssm.edu

      Published 28 October 2015, Sci. Transl. Med. 7, 311ra174 (2015)
      DOI: 10.1126/scitranslmed.aaa9364

      This PDF file includes:

      • Fig. S1. Age distributions for overall, female, and male populations.
      • Table S1. Clinical features.
      • Table S2. Patient characteristics across entire Biobank cohort.
      • Table S3. Significant SNPs specific for each T2D subtype.
      • Table S4. Genes and variants.

      [Download PDF]

    Navigate This Article