Protein Comparison via Model Overlaying using TM Align

backend/.gitignore

@@ -1,3 +1,4 @@
 __pycache__/
 .venv
-foldseek/
+foldseek/
+tmalign/

backend/src/api/protein.py

@@ -5,9 +5,11 @@
 from Bio.PDB import PDBParser
 from Bio.SeqUtils import molecular_weight, seq1
 from ..db import Database, bytea_to_str, str_to_bytea
+from fastapi.exceptions import HTTPException
 
 from ..api_types import ProteinEntry, UploadBody, UploadError, EditBody, CamelModel
 from ..auth import requiresAuthentication
+from ..tmalign import tm_align
 from io import BytesIO
 from fastapi import APIRouter
 from fastapi.responses import FileResponse, StreamingResponse
@@ -335,3 +337,18 @@ def edit_protein_entry(body: EditBody, req: Request):
 
  except Exception:
  return UploadError.WRITE_ERROR
+
+
+# /pdb with two attributes returns both PDBs, superimposed and with different colors.
+@router.get("/protein/pdb/{proteinA:str}/{proteinB:str}")
+def search_proteins(proteinA: str, proteinB: str):
+ try:
+ pdbA = stored_pdb_file_name(proteinA)
+ pdbB = stored_pdb_file_name(proteinB)
+
+ file = tm_align(proteinA, pdbA, proteinB, pdbB)
+
+ return FileResponse(file, filename=proteinA + "_" + proteinB + ".pdb")
+ except Exception as e:
+ log.error(e)
+ raise HTTPException(status_code=500, detail=str(e))

backend/src/api_types.py

@@ -76,3 +76,12 @@ class LoginBody(CamelModel):
 class LoginResponse(CamelModel):
  token: str
  error: str
+
+
+class CompareBody(CamelModel):
+ proteinA: str
+ proteinB: str
+
+
+class CompareResponse(CamelModel):
+ file: list[str]

backend/src/tmalign.py

@@ -0,0 +1,75 @@
+import subprocess
+import logging as log
+import os
+
+
+def bash_cmd(cmd: str | list[str]) -> str:
+ return subprocess.check_output(cmd, shell=True).decode()
+
+
+TMALIGN_LOCATION = "/app/tmalign"
+TMALIGN_EXECUTABLE = f"{TMALIGN_LOCATION}/tmalign"
+
+
+def assert_tmalign_installed():
+ if os.path.exists(TMALIGN_EXECUTABLE):
+ return
+ else:
+ raise ImportError(
+ "tm align executable not installed. Please install manually - Automatic install TODO."
+ )
+
+
+def parse_pdb(filepath: str) -> list[str]:
+ with open(filepath, "r") as f:
+ lines = f.readlines()
+ return lines
+
+
+def tm_align(
+ protein_A: str, pdbA: str, protein_B: str, pdbB: str, type: str = "_all_atm"
+):
+ """
+ Description:
+ Returns two overlaid, aligned, and colored PDB structures in a single PDB file.
+ The ones without extensions appear to be PDB files.
+
+ Params:
+ protein_A:
+ The name of the first protein.
+ pdbA:
+ The filepath of the first protein.
+ protein_B:
+ The name of the second protein.
+ pdbB:
+ The filepath of the second protein.
+ type:
+ The kind of file you want. Experiment with these! Defaults to _all_atm,
+ which shows alpha helices and beta sheets. Valid options include:
+ "", "_all", "_all_atm", "_all_atm_lig", "_atm",
+ ".pml", "_all.pml", "_all_atm.pml", "all_atm_lig.pml", "_atm.pml"
+ """
+ dir_name = protein_A + "-" + protein_B
+ full_path = f"{TMALIGN_LOCATION}/{dir_name}"
+ out_file = full_path + "/output"
+ desired_file = out_file + type
+
+ # If the directory already exists, then we've already run TM align for this protein pair. We can just return the file.
+ if os.path.exists(full_path):
+ log.warn(f"Path {full_path} already exists. Do not need to run TM align.")
+
+ # If the directory doesn't exist, then we need to run TM align and generate the files.
+ else:
+ log.warn(f"Path {full_path} does not exist. Creating directory and returning.")
+ cmd = f"{TMALIGN_EXECUTABLE} {pdbA} {pdbB} -o {out_file}"
+ try:
+ bash_cmd(f"mkdir {full_path}")
+ log.warn(f"Attempting to align now with cmd {cmd}")
+ stdout = bash_cmd(cmd)
+ log.warn(stdout)
+
+ except Exception as e:
+ log.warn(e)
+ raise e
+
+ return desired_file

frontend/src/Router.svelte

@@ -10,6 +10,7 @@
  import Edit from "./routes/Edit.svelte";
  import Tutorials from "./routes/Tutorials.svelte";
  import ForceUploadThumbnails from "./routes/ForceUploadThumbnails.svelte";
+ import Compare from "./routes/Compare.svelte";
 </script>
 
 <Router>
@@ -27,6 +28,7 @@
  >
  <Route path="/edit/:id" let:params><Edit urlId={params.id} /></Route>
  <Route path="/force-upload-thumbnails"><ForceUploadThumbnails /></Route>
+ <Route path="/compare/:a/:b" let:params><Compare proteinA={params.a} proteinB={params.b}/></Route>
  <Route path="/*"><Error /></Route>
  </main>
 </Router>

frontend/src/lib/SimilarProteins.svelte

@@ -28,6 +28,7 @@
  <th> Probability Match</th>
  <th> E-Value </th>
  <th> Description </th>
+ <th> Compare </th>
  </tr>
  {#each similarProteins as protein}
  <tr class="pdb-row">
@@ -41,6 +42,13 @@
  <td>{protein.prob}</td>
  <td>{protein.evalue}</td>
  <td class="pdb-desc">{protein.description}</td>
+ <td>
+ <a
+ use:link
+ href="/compare/{queryProteinName}/{protein.name}"
+ ><LinkOutline size="sm" />Compare</a
+ >
+ </td>
  </tr>
  {/each}
  </table>

frontend/src/routes/Compare.svelte

@@ -0,0 +1,90 @@
+<script lang="ts">
+ import { onMount } from "svelte";
+ import { Backend, BACKEND_URL, type ProteinEntry } from "../lib/backend";
+ import ProteinVis from "../lib/ProteinVis.svelte";
+ import { Button, Dropdown, DropdownItem } from "flowbite-svelte";
+ import Markdown from "../lib/Markdown.svelte";
+ import { numberWithCommas, undoFormatProteinName } from "../lib/format";
+ import { navigate } from "svelte-routing";
+ import References from "../lib/References.svelte";
+ import { ChevronDownSolid, PenOutline } from "flowbite-svelte-icons";
+ import EntryCard from "../lib/EntryCard.svelte";
+ import SimilarProteins from "../lib/SimilarProteins.svelte";
+ import DelayedSpinner from "../lib/DelayedSpinner.svelte";
+ import { user } from "../lib/stores/user";
+
+ const fileDownloadDropdown = ["pdb", "fasta"];
+
+ export let proteinA: string;
+ export let proteinB: string;
+ let combined = proteinA + "/" + proteinB
+ let urlId="Gh_comp1045_c0_seq1"
+ let entry: ProteinEntry | null = null;
+ let error = false;
+
+ // when this component mounts, request protein wikipedia entry from backend
+ onMount(async () => {
+ // Request the protein from backend given ID
+ console.log("Requesting", proteinA, "and", proteinB, "info from backend");
+
+ entry = await Backend.getProteinEntry(urlId);
+ // if we could not find the entry, the id is garbo
+ if (entry == null) error = true;
+ });
+</script>
+
+<svelte:head>
+ <title>Venome Protein {entry ? entry.name : ""}</title>
+</svelte:head>
+
+<section class="flex gap-10 p-5">
+ {#if entry}
+ <div id="left-side">
+ <!-- TITLE AND DESCRIPTION -->
+ <h1 id="title">
+ Comparing Proteins
+ </h1>
+
+ <div id="description">
+ {proteinA} and {proteinB}
+ </div>
+ <ProteinVis
+ format="pdb"
+ proteinName={combined}
+ width={750}
+ height={500}
+ on:mount={async ({ detail: { screenshot } }) => {
+ // upload the protein thumbnail if it doesn't exist
+ if (entry !== null && entry.thumbnail === null) {
+ const b64 = await screenshot();
+ const res = await Backend.uploadProteinPng({
+ proteinName: entry.name,
+ base64Encoding: b64,
+ });
+ }
+ }}
+ />
+ </div>
+ {:else if !error}
+ <!-- Otherwise, tell user we tell the user we are loading -->
+ <h1><DelayedSpinner text="Loading Protein Entry" /></h1>
+ {:else if error}
+ <!-- if we error out, tell the user the id is shiza -->
+ <h1>Error</h1>
+ <p>Could not find a protein with the id <code>{urlId}</code></p>
+ {/if}
+</section>
+
+<style>
+ #left-side {
+ width: 100%;
+ }
+ #right-side {
+ width: 450px;
+ }
+ #title {
+ font-size: 2.45rem;
+ font-weight: 500;
+ color: var(--darkblue);
+ }
+</style>

run.sh

@@ -115,6 +115,20 @@ function remove_foldseek() {
  docker exec -it venome-backend rm -fr foldseek/
 }
 
+function add_tmalign() {
+ docker exec -it venome-backend wget https://seq2fun.dcmb.med.umich.edu//TM-align/TMalign_cpp.gz
+ docker exec -it venome-backend mkdir tmalign
+ docker exec -it venome-backend gzip -d TMalign_cpp.gz
+ docker exec -it venome-backend mv TMalign_cpp tmalign/tmalign
+ docker exec -it venome-backend chmod a+x tmalign/tmalign
+ docker exec -it venome-backend rm -f TMalign_cpp.gz
+}
+
+function remove_tmalign() {
+ docker exec -it venome-backend rm -f TMalign_cpp.gz*
+ docker exec -it venome-backend rm -fr tmalign/
+}
+
 function scrape_func_names() {
  functions=($(grep -oE 'function[[:space:]]+[a-zA-Z_][a-zA-Z_0-9]*' ./run.sh | sed 's/function[[:space:]]*//'))
 }