Skip to main content

TP53 from prediction to crystal

TL;DR: the p53 protein has a predicted structure covering every residue and crystal structures covering the parts that fold. This page opens three of them in one view beside the TP53 gene, superposed and each mapped to the same transcript, then clicks a cancer hotspot on the crystal and reads it back to its codon and its ClinVar entries. The protein3d plugin does the mapping; Mol* draws the structures.

Prerequisites

  • nothing to install: every link below opens a hosted JBrowse that already loads the protein3d plugin, and the structures, annotations and mappings are fetched live from the services named next

Where the data comes from

The hg38 config the links open carries NCBI RefSeq and ClinVar; everything about the protein arrives per structure from the services beside it.

Three structures of one protein

An AlphaFold model is one chain, numbered like the UniProt sequence it was predicted from. A crystal structure is whatever was crystallised: a domain cut out of the protein, sometimes several copies of it, often with a partner or a piece of DNA, and numbered from wherever the construct began. Putting the two kinds beside a gene means answering the same question for each, which residue of the structure is which codon of the transcript, and the plugin answers it by aligning each structure's own sequence to the transcript's translation.

Open the three structures of TP53. The link is a session spec naming the gene's locus, its RefSeq transcript NM_000546.6, and three structures by id: a UniProt accession for the AlphaFold model and two PDB ids. The plugin resolves each id to a file, translates the transcript's CDS against hg38, aligns every structure to that translation, and superposes the structures with TM-align.

TP53 on hg38 with NCBI RefSeq and ClinVar beside a protein view holding the AlphaFold model, 1TUP and 1YCR. Each structure has its own alignment panel over the shared Mol* canvas, where the two crystal fragments sit on the model's core and the model's tails run on past them. Open this view in JBrowse ↗ (large dataset — this session takes a while to load)

The steps behind the figure above, run against your own data.

    Some settings have no written step yet — see the Session spec tab.

    Open this view in JBrowse Desktop ↗

    Opens JBrowse Desktop (JBrowse Desktop 5.0+) at this view and saves it as a reopenable session — swap in your own files afterwards.

    Nothing happens? Paste this link into Desktop's Open JBrowse Web link... (start screen, or File → Session):

    https://jbrowse.org/code/jb2/main/?config=test_data%2Fprotein3d_config.json&session=spec-%7B%22views%22%3A%5B%7B%22type%22%3A%22ProteinView%22%2C%22structures%22%3A%5B%7B%22uniprotId%22%3A%22P04637%22%7D%2C%7B%22pdbId%22%3A%221TUP%22%7D%2C%7B%22pdbId%22%3A%221YCR%22%7D%5D%2C%22transcriptId%22%3A%22NM_000546.6%22%2C%22height%22%3A540%2C%22sideBySide%22%3Atrue%2C%22zoomToBaseLevel%22%3Afalse%2C%22connectedView%22%3A%7B%22assembly%22%3A%22hg38%22%2C%22loc%22%3A%22chr17%3A7%2C671%2C000-7%2C684%2C500%22%2C%22tracks%22%3A%5B%7B%22trackId%22%3A%22hg38-ncbiRefSeq%22%2C%22height%22%3A150%7D%2C%7B%22trackId%22%3A%22clinvar_ncbi_hg38%22%2C%22height%22%3A1040%7D%5D%7D%7D%5D%7D&sessionName=JBrowse+docs%3A+protein%2Ftp53_three_structures

    This session spec pastes after &session=spec- on any JBrowse Web instance.

    {
  "views": [
    {
      "type": "ProteinView",
      "structures": [
        {
          "uniprotId": "P04637"
        },
        {
          "pdbId": "1TUP"
        },
        {
          "pdbId": "1YCR"
        }
      ],
      "transcriptId": "NM_000546.6",
      "height": 540,
      "sideBySide": true,
      "zoomToBaseLevel": false,
      "connectedView": {
        "assembly": "hg38",
        "loc": "chr17:7,671,000-7,684,500",
        "tracks": [
          {
            "trackId": "hg38-ncbiRefSeq",
            "height": 150
          },
          {
            "trackId": "clinvar_ncbi_hg38",
            "height": 1040
          }
        ]
      }
    }
  ]
}

    Loaded against config: test_data/protein3d_config.json

    Hand this to a coding agent. It rebuilds the figure above headlessly, and the session file is then the thing to edit — swap an adapter uri for your own file and rerun.

    cat > session.json <<'JSON'
{
  "views": [
    {
      "type": "ProteinView",
      "structures": [
        {
          "uniprotId": "P04637"
        },
        {
          "pdbId": "1TUP"
        },
        {
          "pdbId": "1YCR"
        }
      ],
      "transcriptId": "NM_000546.6",
      "height": 540,
      "sideBySide": true,
      "zoomToBaseLevel": false,
      "connectedView": {
        "assembly": "hg38",
        "loc": "chr17:7,671,000-7,684,500",
        "tracks": [
          {
            "trackId": "hg38-ncbiRefSeq",
            "height": 150
          },
          {
            "trackId": "clinvar_ncbi_hg38",
            "height": 1040
          }
        ]
      }
    }
  ]
}
JSON

npx @jbrowse/capture --instance https://jbrowse.org/code/jb2/main/ \
  --config https://jbrowse.org/code/jb2/main/test_data/protein3d_config.json \
  --session session.json -o figure.png

    Using JBrowse with AI agents covers the rest: hosted genomes to point it at, and why waiting for the render is the part that goes wrong.

    Each alignment panel puts the transcript's translation on the GENOME row and the structure's own sequence on the STRUCT row, with a residue ruler under them in the numbering the structure's authors assigned. The AlphaFold panel is one unbroken match. The 1TUP panel starts in the middle of the transcript and stops well before its end, because the crystallised construct is the DNA-binding core; its ruler starts at 94, the residue of p53 the construct begins at, so a tick under the row is the number a paper would cite.

    Under the STRUCT row of the AlphaFold panel, pLDDT is high across the core and falls away at both ends of the protein, which is the same region the crystal panels leave as gap: the tails that the crystals do not contain are the tails the model is least sure of.

    The crystal keeps its annotations

    The feature tracks under each panel come from UniProt, whose coordinates are the full-length sequence. For the AlphaFold model that is the structure's own numbering. For 1TUP the plugin asks SIFTS where the construct starts and shifts every feature by that offset, so the DNA binding region and the natural variants land on the residues the crystal actually has, and features outside the construct are dropped. The caption under the panel names the UniProt entry the tracks came from.

    A crystal can also hold more than one chain. 1TUP has three copies of the core and two DNA strands, and the panel's Mapped chain picker lists them; the plugin chose the protein entity because it is the one whose sequence aligns to the transcript. Hover any of the three copies in the 3D canvas and the same residue lights on the genome.

    Click a hotspot

    Open the same session with R248 selected, or click residue 248 on the 1TUP panel's STRUCT row yourself: the ruler under the row and the transcript row above agree on the number, because 1TUP's authors numbered their construct the way UniProt numbers the whole protein.

    R248 selected on 1TUP: magenta on the crystal's DNA-contact loop, a band over its codon on NCBI RefSeq, and the ClinVar rows that fall inside the band. The AlphaFold model carries the same residue, so the hover reaches both structures. Open this view in JBrowse ↗ (large dataset — this session takes a while to load)

    The steps behind the figure above, run against your own data.

      Some settings have no written step yet — see the Session spec tab.

      Open this view in JBrowse Desktop ↗

      Opens JBrowse Desktop (JBrowse Desktop 5.0+) at this view and saves it as a reopenable session — swap in your own files afterwards.

      Nothing happens? Paste this link into Desktop's Open JBrowse Web link... (start screen, or File → Session):

      https://jbrowse.org/code/jb2/main/?config=test_data%2Fprotein3d_config.json&session=spec-%7B%22views%22%3A%5B%7B%22type%22%3A%22ProteinView%22%2C%22structures%22%3A%5B%7B%22uniprotId%22%3A%22P04637%22%7D%2C%7B%22pdbId%22%3A%221TUP%22%2C%22initialResidues%22%3A%7B%22start%22%3A248%2C%22end%22%3A248%7D%7D%2C%7B%22pdbId%22%3A%221YCR%22%7D%5D%2C%22transcriptId%22%3A%22NM_000546.6%22%2C%22height%22%3A540%2C%22sideBySide%22%3Atrue%2C%22zoomToBaseLevel%22%3Afalse%2C%22connectedView%22%3A%7B%22assembly%22%3A%22hg38%22%2C%22loc%22%3A%22chr17%3A7%2C674%2C100-7%2C674%2C350%22%2C%22tracks%22%3A%5B%7B%22trackId%22%3A%22hg38-ncbiRefSeq%22%2C%22height%22%3A150%7D%2C%7B%22trackId%22%3A%22clinvar_ncbi_hg38%22%2C%22height%22%3A1040%7D%5D%7D%7D%5D%7D&sessionName=JBrowse+docs%3A+protein%2Ftp53_hotspot

      This session spec pastes after &session=spec- on any JBrowse Web instance.

      {
  "views": [
    {
      "type": "ProteinView",
      "structures": [
        {
          "uniprotId": "P04637"
        },
        {
          "pdbId": "1TUP",
          "initialResidues": {
            "start": 248,
            "end": 248
          }
        },
        {
          "pdbId": "1YCR"
        }
      ],
      "transcriptId": "NM_000546.6",
      "height": 540,
      "sideBySide": true,
      "zoomToBaseLevel": false,
      "connectedView": {
        "assembly": "hg38",
        "loc": "chr17:7,674,100-7,674,350",
        "tracks": [
          {
            "trackId": "hg38-ncbiRefSeq",
            "height": 150
          },
          {
            "trackId": "clinvar_ncbi_hg38",
            "height": 1040
          }
        ]
      }
    }
  ]
}

      Loaded against config: test_data/protein3d_config.json

      Hand this to a coding agent. It rebuilds the figure above headlessly, and the session file is then the thing to edit — swap an adapter uri for your own file and rerun.

      cat > session.json <<'JSON'
{
  "views": [
    {
      "type": "ProteinView",
      "structures": [
        {
          "uniprotId": "P04637"
        },
        {
          "pdbId": "1TUP",
          "initialResidues": {
            "start": 248,
            "end": 248
          }
        },
        {
          "pdbId": "1YCR"
        }
      ],
      "transcriptId": "NM_000546.6",
      "height": 540,
      "sideBySide": true,
      "zoomToBaseLevel": false,
      "connectedView": {
        "assembly": "hg38",
        "loc": "chr17:7,674,100-7,674,350",
        "tracks": [
          {
            "trackId": "hg38-ncbiRefSeq",
            "height": 150
          },
          {
            "trackId": "clinvar_ncbi_hg38",
            "height": 1040
          }
        ]
      }
    }
  ]
}
JSON

npx @jbrowse/capture --instance https://jbrowse.org/code/jb2/main/ \
  --config https://jbrowse.org/code/jb2/main/test_data/protein3d_config.json \
  --session session.json -o figure.png

      Using JBrowse with AI agents covers the rest: hosted genomes to point it at, and why waiting for the render is the part that goes wrong.

      The selection is magenta on the crystal, where the residue sits in the loop that reaches into the DNA's minor groove, and a band on the genome tracks over the codon. The ClinVar rows inside the band are the substitutions at that codon. Hover one of them and the residue lights on both structures at once, since both map it; hover the intron beside the exon and nothing lights anywhere.

      The complex maps the right chain

      1YCR is not a p53 structure so much as an MDM2 structure with a piece of p53 in it: MDM2's N-terminal domain and a fifteen-residue peptide from p53's transactivation region. Both are chains of the file, and only one of them is encoded by the transcript.

      The 1YCR panel's Mapped chain picker shows both. The plugin picked the peptide, whose alignment is a short exact match near the start of the transcript row; MDM2 is listed above it.

      1YCR's Mapped chain picker open. The p53 peptide, Chain B, is the mapped one; Chain A is MDM2. Open this view in JBrowse ↗ (large dataset — this session takes a while to load)

      The steps behind the figure above, run against your own data.

        Some settings have no written step yet — see the Session spec tab.

        Open this view in JBrowse Desktop ↗

        Opens JBrowse Desktop (JBrowse Desktop 5.0+) at this view and saves it as a reopenable session — swap in your own files afterwards.

        Nothing happens? Paste this link into Desktop's Open JBrowse Web link... (start screen, or File → Session):

        https://jbrowse.org/code/jb2/main/?config=test_data%2Fprotein3d_config.json&session=spec-%7B%22views%22%3A%5B%7B%22type%22%3A%22ProteinView%22%2C%22structures%22%3A%5B%7B%22uniprotId%22%3A%22P04637%22%7D%2C%7B%22pdbId%22%3A%221TUP%22%7D%2C%7B%22pdbId%22%3A%221YCR%22%7D%5D%2C%22transcriptId%22%3A%22NM_000546.6%22%2C%22height%22%3A540%2C%22sideBySide%22%3Atrue%2C%22zoomToBaseLevel%22%3Afalse%2C%22connectedView%22%3A%7B%22assembly%22%3A%22hg38%22%2C%22loc%22%3A%22chr17%3A7%2C671%2C000-7%2C684%2C500%22%2C%22tracks%22%3A%5B%7B%22trackId%22%3A%22hg38-ncbiRefSeq%22%2C%22height%22%3A150%7D%2C%7B%22trackId%22%3A%22clinvar_ncbi_hg38%22%2C%22height%22%3A1040%7D%5D%7D%7D%5D%7D&sessionName=JBrowse+docs%3A+protein%2Ftp53_mapped_chain

        This session spec pastes after &session=spec- on any JBrowse Web instance.

        {
  "views": [
    {
      "type": "ProteinView",
      "structures": [
        {
          "uniprotId": "P04637"
        },
        {
          "pdbId": "1TUP"
        },
        {
          "pdbId": "1YCR"
        }
      ],
      "transcriptId": "NM_000546.6",
      "height": 540,
      "sideBySide": true,
      "zoomToBaseLevel": false,
      "connectedView": {
        "assembly": "hg38",
        "loc": "chr17:7,671,000-7,684,500",
        "tracks": [
          {
            "trackId": "hg38-ncbiRefSeq",
            "height": 150
          },
          {
            "trackId": "clinvar_ncbi_hg38",
            "height": 1040
          }
        ]
      }
    }
  ]
}

        Loaded against config: test_data/protein3d_config.json

        Hand this to a coding agent. It rebuilds the figure above headlessly, and the session file is then the thing to edit — swap an adapter uri for your own file and rerun.

        cat > session.json <<'JSON'
{
  "views": [
    {
      "type": "ProteinView",
      "structures": [
        {
          "uniprotId": "P04637"
        },
        {
          "pdbId": "1TUP"
        },
        {
          "pdbId": "1YCR"
        }
      ],
      "transcriptId": "NM_000546.6",
      "height": 540,
      "sideBySide": true,
      "zoomToBaseLevel": false,
      "connectedView": {
        "assembly": "hg38",
        "loc": "chr17:7,671,000-7,684,500",
        "tracks": [
          {
            "trackId": "hg38-ncbiRefSeq",
            "height": 150
          },
          {
            "trackId": "clinvar_ncbi_hg38",
            "height": 1040
          }
        ]
      }
    }
  ]
}
JSON

npx @jbrowse/capture --instance https://jbrowse.org/code/jb2/main/ \
  --config https://jbrowse.org/code/jb2/main/test_data/protein3d_config.json \
  --session session.json -o figure.png

        Using JBrowse with AI agents covers the rest: hosted genomes to point it at, and why waiting for the render is the part that goes wrong.

        Switch the picker to Chain A. The alignment is recomputed against MDM2, and the GENOME row becomes a scatter of gapped fragments, since nothing in the transcript encodes it; hovering the structure now reaches no consistent codon. Switch back to Chain B and the peptide's alignment returns. A complex of two paralogs is where the plugin's own choice can land on the wrong one, and the picker is the way back.

        Checking the hotspot against the sequence

        Back on the genome view, zoom into the band the R248 selection drew, down to base level. The codon under the band is CGG; the ClinVar rows over it are the R248Q and R248W entries, one changing the middle base and one the first. The transcript is on the minus strand, so the codon reads right to left on the reference track.

        See also

        References

        • AlphaFold DB
        • RCSB PDB
        • UniProt
        • SIFTS
        • Cho Y, Gorina S, Jeffrey PD, Pavletich NP. Crystal structure of a p53 tumor suppressor-DNA complex: understanding tumorigenic mutations. Science 1994.
        • Kussie PH, Gorina S, Marechal V, et al. Structure of the MDM2 oncoprotein bound to the p53 tumor suppressor transactivation domain. Science 1996.