DPCfam Metaclusters

🧬 Domain Classification in Sequence Space

The DPCfam pipeline applied the Density Peak Clustering (DPC) algorithm to roughly 23 million UniRef50 sequences (UniRef50 v. 2017_07) and automatically grouped homologous domain segments into 81,384 protein families called metaclusters, entirely without manual curation. Each metacluster was then used to build a multiple sequence alignment and a profile-HMM for downstream analysis.

The table below lists all 81,384 DPCfam metaclusters by default. Use the toggle buttons above to switch to the Standard subset (size ≥ 50 seeds, 46,828 metaclusters) or the DPCfamB subset (25 to 49 seeds, 34,556 metaclusters). Intrinsic metacluster properties appear on the left; Pfam annotations on the right. Click any MCID to open its detail page.

MCID Size UniRef50 Avg. Len. Std avg len % LC % CC % DIS Avg. TM Pfam DA % DA % Avg. Ov. Overlap Label
MC184390 76 217.74 19.76 31.48 0.0 2.95 7.08 UNKNOWN 0.0 0.0 NONE
MC184407 53 83.02 7.37 0.59 0.0 16.48 0.0 UNKNOWN 0.0 0.0 NONE
MC184410 89 104.26 12.94 0.64 0.0 22.18 0.03 PF01398 18.33 7.93 Shifted
MC184427 410 83.73 4.15 1.2 0.0 22.88 0.0 PF09524 98.53 85.91 Equivalent
MC184428 135 97.51 18.0 7.68 0.0 27.75 0.01 PF14297 97.67 24.04 Shifted
MC184452 137 104.73 10.37 1.42 0.35 46.45 0.0 UNKNOWN 0.0 0.0 NONE
MC184469 113 55.5 11.67 2.79 7.55 27.75 0.0 PF08499 21.98 3.55 Shifted
MC184470 26 73.08 7.92 0.03 0.0 18.16 0.0 PF09347 100.0 17.16 Shifted
MC184471 669 60.74 6.74 1.21 0.0 20.06 0.01 PF12571 99.74 33.7 Reduced
MC184475 42 121.95 14.96 0.06 14.01 30.63 0.0 UNKNOWN 0.0 0.0 NONE
Legend
  • A. Metacluster Intrinsic Properties:
  • MCID: Unique identifier per metacluster.
  • 1. Size UniRef50 : Number of UniRef50 seed sequences in the metacluster.
  • 2. Avg. Len.: Mean domain length of the seeds (in amino acids).
  • 3. Std. Avg. Len.: Standard deviation of seed domain lengths.
  • 4. % LC Percent of residues located in low-complexity regions.
  • 5. % CC: Percent of residues located in coiled-coil regions.
  • 6. % DIS: Percent of residues predicted as intrinsically disordered.
  • 7. Avg.TM: Average number of transmembrane regions per domain.
  • B. Pfam Annotations:
  • 1. Pfam DA: Pfam architecture which best overlaps with MC. We refer to it as the dominant architecture (DA). If there is no overlap, the MC is tagged as UNKNOWN (representing 'dark matter').
  • 2. % DA: Percent of sequences in the MC matching the DA.
  • 3. % Avg. Ov.: Average % boundary overlap between the MC and Pfam, computed only for seeds that match the DA.
  • 4. Overlap Label: Label describing qualitatively the overlap of the MC with its DA, which can be : equivalent, reduced, extended or shifted. NONE, otherwise.
    Color-coded based on boundary agreement:
    • Yellow (Equivalent): the MC and its dominant architecture share very similar boundaries.
    • Blue (Reduced): the MC covers only part of the dominant architecture.
    • Pink (Extended): the MC extends beyond the dominant architecture boundaries.
    • Green (Shifted): the MC and the dominant architecture share a boundary offset, with poor positional agreement.