GPCR-ligand predictions Machine Learning Case Study

For this case study, we trained a classifier to predict known interactions between G protein-coupled receptors (GPCRs) and protein-derived ligands, and then predicted ligands for orphan GPRCs. The classifier performed well in cross-validation with an area under the receiver operating characteristic curve of 0.91, F1-score of 0.59, and Matthew’s correlation coefficient of 0.59. Methods and results for this machine learning case study are described in detail in the Harmonizome publication.

Each row in the results table provides the gene symbol of a GPCR (GPCR column), provides the gene symbol of a candidate protein-derived ligand (Ligand column), provides the predicted probability that the candidate ligand binds to the GPCR (Probability column), indicates whether pair is known to interact (Known column), indicates whether the pair was used as an example for training the classifier (Training column), and provides the predicted false discovery rate if all predictions with better or equivalent probability were tested (FDR column).

The table below shows predictions with probabilities greater than or equal to 0.5. To download the full table:

GPCR Ligand Probability Known Training FDR
CCR5 CCL7 0.92 true true 0.0
CXCR2 CXCL1 0.89 true true 0.0
LGR4 RSPO4 0.89 true true 0.0
CCR5 CCL8 0.88 true true 0.0
CXCR2 CXCL6 0.87 true true 0.0
LGR4 RSPO1 0.86 true true 0.0
NPY5R NPY 0.8 true true 0.0
LGR4 RSPO2 0.8 true true 0.0
NPY5R PYY 0.79 true true 0.08
LGR5 RSPO4 0.79 true true 0.08
RXFP2 INS 0.79 false false 0.09
CCR1 CCL11 0.79 false true 0.09
LGR6 RSPO4 0.79 true true 0.0
CCR2 CCL7 0.79 true true 0.0
RXFP2 IGF2 0.78 false false 0.07
RXFP2 IGF1 0.77 false false 0.07
CXCR2 PF4V1 0.77 false false 0.07
CCR3 CCL7 0.77 true true 0.07
LGR5 RSPO1 0.76 true true 0.07
CXCR2 CXCL5 0.76 true true 0.06
CCR5 CCL13 0.76 true true 0.06
RXFP1 INS 0.76 false false 0.06
RXFP1 RLN1 0.76 true true 0.06
RXFP2 RLN1 0.75 true true 0.05
CCR3 CCL8 0.75 true true 0.05
RXFP1 IGF1 0.74 false false 0.05
CCR1 CCL5 0.74 true true 0.05
CXCR2 PF4 0.74 false false 0.05
NPY1R PYY 0.73 true true 0.05
FZD6 WNT3 0.73 false false 0.04
NPY1R NPY 0.73 true true 0.04
CCR1 CCL8 0.73 true true 0.04
FZD6 WNT5A 0.72 true true 0.1
CCR5 CCL11 0.72 true true 0.11
RXFP1 IGF2 0.72 false false 0.11
AVPR2 OXT 0.72 true true 0.11
CCR1 CCL16 0.72 false true 0.12
CCR1 CCL2 0.72 false true 0.08
CCR2 CCL5 0.71 false true 0.12
AVPR1B OXT 0.71 true true 0.09
TACR1 TAC3 0.71 true true 0.09
LGR5 RSPO2 0.71 true true 0.1
CCR1 CCL7 0.71 true true 0.1
SSTR4 CORT 0.7 true true 0.11
LGR4 RSPO3 0.7 true true 0.11
CCR1 CCL13 0.69 true true 0.09
CCR5 CCL5 0.69 true true 0.09
SSTR1 SST 0.69 true true 0.1
AVPR1A AVP 0.69 true true 0.1
FZD6 WNT4 0.69 true true 0.1
VIPR1 VIP 0.69 true true 0.1
SSTR1 CORT 0.69 true true 0.11
AVPR1A OXT 0.69 true true 0.11
CCR2 CCL8 0.68 true true 0.09
NPY2R PPY 0.68 false true 0.11
CCR3 CCL3 0.67 false true 0.12
FZD6 WNT3A 0.67 true true 0.1
MC4R ASIP 0.67 true true 0.1
CXCR3 CXCL9 0.67 true true 0.11
RXFP2 RLN3 0.66 true true 0.14
CCR5 CCL23 0.66 false true 0.14
SSTR5 SST 0.66 true true 0.13
SSTR5 CORT 0.66 true true 0.13
CCR5 CCL15 0.66 false true 0.13
CXCR1 CXCL8 0.66 true true 0.11
CXCR2 CXCL8 0.66 true true 0.12
LGR5 RSPO3 0.66 true true 0.12
VIPR1 GIP 0.65 false true 0.15
VIPR1 SCT 0.65 true true 0.15
CCR5 CCL3 0.64 true true 0.15
SSTR4 SST 0.63 true true 0.15
VIPR1 GCG 0.63 false true 0.16
OXTR OXT 0.63 true true 0.16
CXCR2 CXCL3 0.62 true true 0.15
LGR6 RSPO1 0.62 true true 0.15
CCR2 CCL13 0.62 true true 0.15
CXCR2 YARS 0.62 false true 0.16
CXCR3 CXCL10 0.61 true true 0.15
AVPR1B AVP 0.61 true true 0.15
AVPR2 AVP 0.61 true true 0.16
FPR3 ANXA1 0.61 true true 0.16
OPRL1 POMC 0.6 false true 0.16
SSTR2 SST 0.6 true true 0.16
VIPR2 SCT 0.6 true true 0.16
CCR5 CCL2 0.6 true true 0.16
CCR3 CCL16 0.6 false true 0.17
MC3R ASIP 0.6 true true 0.17
CCR1 CCL3 0.59 true true 0.16
NPY2R NPY 0.59 true true 0.16
VIPR2 GCG 0.58 false true 0.16
MC5R ASIP 0.58 true true 0.16
CCR2 CCL11 0.58 true true 0.15
RXFP1 RLN3 0.58 true true 0.15
OPRD1 PDYN 0.58 true true 0.15
CCR3 CCL13 0.58 true true 0.16
SSTR3 CORT 0.58 true true 0.16
SSTR2 CORT 0.58 true true 0.16
OPRK1 PDYN 0.57 true true 0.16
CXCR3 CCL8 0.57 false true 0.16
CCR2 CCL2 0.57 true true 0.15
NPY5R PPY 0.57 true true 0.16
TACR3 TAC3 0.57 true true 0.16
NPY2R PYY 0.56 true true 0.16
MC5R AGRP 0.56 true true 0.16
TACR2 TAC3 0.56 true true 0.16
NPY1R PPY 0.56 true true 0.17
ADCYAP1R1 GCG 0.56 false true 0.17
CCR3 CCL11 0.56 true true 0.16
RXFP2 INSL5 0.55 false true 0.17
LGR6 RSPO2 0.55 true true 0.17
SSTR3 SST 0.55 true true 0.17
OPRM1 PDYN 0.55 true true 0.17
VIPR2 VIP 0.54 true true 0.16
RXFP3 IGF2 0.54 false false 0.16
CCR1 CCL18 0.54 false false 0.16
NPBWR1 NPW 0.54 true true 0.16
CXCR2 PPBP 0.54 true true 0.16
OPRK1 PENK 0.53 true true 0.18
LGR6 RSPO3 0.53 true true 0.18
CXCR1 PPBP 0.53 false true 0.18
MRGPRX2 AGT 0.53 false true 0.17
NPBWR1 NPB 0.53 true true 0.17
FPR3 CAMP 0.53 false true 0.17
CCR3 CCL2 0.53 true true 0.16
RXFP3 INS 0.53 false false 0.16
OPRD1 PENK 0.52 true true 0.19
CCR2 CCL3 0.52 false true 0.19
CCR5 CCL18 0.52 false false 0.19
GLP1R SCT 0.52 false true 0.18
CXCR1 PF4V1 0.52 false false 0.18
CCR2 CCL14 0.51 false true 0.19
ADCYAP1R1 VIP 0.51 true true 0.18
CCR5 CCL14 0.51 true true 0.18
MC4R AGRP 0.51 true true 0.19
MC3R POMC 0.51 true true 0.19
OPRM1 PENK 0.5 true true 0.19
OXTR AVP 0.5 true true 0.19
CXCR3 CCL7 0.5 true true 0.19
Funding