app/fn.py

from datetime import datetime
from email.mime.multipart import MIMEMultipart
from email.mime.text import MIMEText
from email.utils import formatdate, make_msgid
from functools import cache
import os
from pathlib import Path
import smtplib
import sys
import tempfile

import pandas as pd
from bokeh.models import NumberFormatter, BooleanFormatter, HTMLTemplateFormatter
import gradio as gr
import pytz
import panel as pn
import seaborn as sns
from markdown import markdown
from rdkit import Chem, RDConfig
from rdkit.Chem import PandasTools, Crippen, Descriptors, rdMolDescriptors, Lipinski, rdmolops
import requests

from app import static

sys.path.append(str(Path(RDConfig.RDContribDir) / 'SA_Score'))
import sascorer


def lipinski(mol):
    """
    Lipinski's rules:
    Hydrogen bond donors <= 5
    Hydrogen bond acceptors <= 10
    Molecular weight <= 500 daltons
    logP <= 5
    """
    return (
            Lipinski.NumHDonors(mol) <= 5 and
            Lipinski.NumHAcceptors(mol) <= 10 and
            Descriptors.MolWt(mol) <= 500 and
            Crippen.MolLogP(mol) <= 5
    )


def reos(mol):
    """
    Rapid Elimination Of Swill filter:
    Molecular weight between 200 and 500
    LogP between -5.0 and +5.0
    H-bond donor count between 0 and 5
    H-bond acceptor count between 0 and 10
    Formal charge between -2 and +2
    Rotatable bond count between 0 and 8
    Heavy atom count between 15 and 50
    """
    return (
            200 <= Descriptors.MolWt(mol) <= 500 and
            -5.0 <= Crippen.MolLogP(mol) <= 5.0 and
            0 <= Lipinski.NumHDonors(mol) <= 5 and
            0 <= Lipinski.NumHAcceptors(mol) <= 10 and
            -2 <= rdmolops.GetFormalCharge(mol) <= 2 and
            0 <= rdMolDescriptors.CalcNumRotatableBonds(mol) <= 8 and
            15 <= rdMolDescriptors.CalcNumHeavyAtoms(mol) <= 50
    )


def ghose(mol):
    """
    Ghose drug like filter:
    Molecular weight between 160 and 480
    LogP between -0.4 and +5.6
    Atom count between 20 and 70
    Molar refractivity between 40 and 130
    """
    return (
            160 <= Descriptors.MolWt(mol) <= 480 and
            -0.4 <= Crippen.MolLogP(mol) <= 5.6 and
            20 <= rdMolDescriptors.CalcNumAtoms(mol) <= 70 and
            40 <= Crippen.MolMR(mol) <= 130
    )


def veber(mol):
    """
    The Veber filter is a rule of thumb filter for orally active drugs described in
    Veber et al., J Med Chem. 2002; 45(12): 2615-23.:
    Rotatable bonds <= 10
    Topological polar surface area <= 140
    """
    return (
            rdMolDescriptors.CalcNumRotatableBonds(mol) <= 10 and
            rdMolDescriptors.CalcTPSA(mol) <= 140
    )


def rule_of_three(mol):
    """
    Rule of Three filter (Congreve et al., Drug Discov. Today. 8 (19): 876–7, (2003).):
    Molecular weight <= 300
    LogP <= 3
    H-bond donor <= 3
    H-bond acceptor count <= 3
    Rotatable bond count <= 3
    """
    return (
            Descriptors.MolWt(mol) <= 300 and
            Crippen.MolLogP(mol) <= 3 and
            Lipinski.NumHDonors(mol) <= 3 and
            Lipinski.NumHAcceptors(mol) <= 3 and
            rdMolDescriptors.CalcNumRotatableBonds(mol) <= 3
    )


@cache
def load_smarts_patterns(smarts_path):
    # Load the CSV file containing SMARTS patterns
    smarts_df = pd.read_csv(Path(smarts_path))
    # Convert all SMARTS patterns to molecules
    smarts_mols = [Chem.MolFromSmarts(smarts) for smarts in smarts_df['smarts']]
    return smarts_mols


def smarts_filter(mol, smarts_mols):
    for smarts_mol in smarts_mols:
        if smarts_mol is not None and mol.HasSubstructMatch(smarts_mol):
            return False
    return True


def pains(mol):
    smarts_mols = load_smarts_patterns("data/filters/pains.csv")
    return smarts_filter(mol, smarts_mols)


def mlsmr(mol):
    smarts_mols = load_smarts_patterns("data/filters/mlsmr.csv")
    return smarts_filter(mol, smarts_mols)


def dundee(mol):
    smarts_mols = load_smarts_patterns("data/filters/dundee.csv")
    return smarts_filter(mol, smarts_mols)


def glaxo(mol):
    smarts_mols = load_smarts_patterns("data/filters/glaxo.csv")
    return smarts_filter(mol, smarts_mols)


def bms(mol):
    smarts_mols = load_smarts_patterns("data/filters/bms.csv")
    return smarts_filter(mol, smarts_mols)


SCORE_MAP = {
    'SAscore': sascorer.calculateScore,
    'LogP': Crippen.MolLogP,
    'Molecular Weight': Descriptors.MolWt,
    'Number of Atoms': rdMolDescriptors.CalcNumAtoms,
    'Number of Heavy Atoms': rdMolDescriptors.CalcNumHeavyAtoms,
    'Molar Refractivity': Crippen.MolMR,
    'H-Bond Donor Count': Lipinski.NumHDonors,
    'H-Bond Acceptor Count': Lipinski.NumHAcceptors,
    'Rotatable Bond Count': rdMolDescriptors.CalcNumRotatableBonds,
    'Topological Polar Surface Area': rdMolDescriptors.CalcTPSA,
}

FILTER_MAP = {
    # TODO support number_of_violations
    'REOS': reos,
    "Lipinski's Rule of Five": lipinski,
    'Ghose': ghose,
    'Rule of Three': rule_of_three,
    'Veber': veber,
    'PAINS': pains,
    'MLSMR': mlsmr,
    'Dundee': dundee,
    'Glaxo': glaxo,
    'BMS': bms,
}


def validate_columns(df, mandatory_cols):
    missing_cols = [col for col in mandatory_cols if col not in df.columns]
    if missing_cols:
        error_message = (f"The following mandatory columns are missing "
                         f"in the uploaded dataset: {str(mandatory_cols).strip('[]')}.")
        raise ValueError(error_message)
    else:
        return


def get_timezone_by_ip(ip, session):
    try:
        data = session.get(f'https://worldtimeapi.org/api/ip/{ip}').json()
        return data['timezone']
    except Exception:
        return 'UTC'


def ts_to_str(timestamp, timezone):
    # Create a timezone-aware datetime object from the UNIX timestamp
    dt = datetime.fromtimestamp(timestamp, pytz.utc)

    # Convert the timezone-aware datetime object to the target timezone
    target_timezone = pytz.timezone(timezone)
    localized_dt = dt.astimezone(target_timezone)

    # Format the datetime object to the specified string format
    return localized_dt.strftime('%Y-%m-%d %H:%M:%S (%Z%z)')


def send_email(job_info):
    if job_info.get('email'):
        try:
            email_info = job_info.copy()
            email_serv = os.getenv('EMAIL_SERV')
            email_port = os.getenv('EMAIL_PORT')
            email_addr = os.getenv('EMAIL_ADDR')
            email_pass = os.getenv('EMAIL_PASS')
            email_form = os.getenv('EMAIL_FORM')
            email_subj = os.getenv('EMAIL_SUBJ')

            for key, value in email_info.items():
                if key.endswith("time") and value:
                    email_info[key] = ts_to_str(value, get_timezone_by_ip(email_info['ip']))

            server = smtplib.SMTP(email_serv, int(email_port))
            # server.starttls()

            server.login(email_addr, email_pass)
            msg = MIMEMultipart("alternative")
            msg["From"] = email_addr
            msg["To"] = email_info['email']
            msg["Subject"] = email_subj.format(**email_info)
            msg["Date"] = formatdate(localtime=True)
            msg["Message-ID"] = make_msgid()

            msg.attach(MIMEText(markdown(email_form.format(**email_info)), 'html'))
            msg.attach(MIMEText(email_form.format(**email_info), 'plain'))

            server.sendmail(email_addr, email_info['email'], msg.as_string())
            server.quit()
            gr.Info('Email notification sent.')
        except Exception as e:
            gr.Warning('Failed to send email notification due to error: ' + str(e))


def read_molecule(path):
    if path.endswith('.pdb'):
        return Chem.MolFromPDBFile(path, sanitize=False, removeHs=True)
    if path.endswith('.pdr'):
        return open(path, 'r').read()
    elif path.endswith('.mol'):
        return Chem.MolFromMolFile(path, sanitize=False, removeHs=True)
    elif path.endswith('.mol2'):
        return Chem.MolFromMol2File(path, sanitize=False, removeHs=True)
    elif path.endswith('.sdf'):
        return Chem.SDMolSupplier(path, sanitize=False, removeHs=True)[0]
    raise Exception('Unknown file extension')


def read_molecule_file(in_file, allowed_extentions):
    if isinstance(in_file, str):
        path = in_file
    else:
        path = in_file.name
    extension = path.split('.')[-1]

    if extension not in allowed_extentions:
        msg = static.INVALID_FORMAT_MSG.format(extension=extension)
        return None, None, msg

    try:
        mol = read_molecule(path)
    except Exception as e:
        e = str(e).replace('\'', '')
        msg = static.ERROR_FORMAT_MSG.format(message=e)
        return None, None, msg

    if extension in 'pdb':
        content = Chem.MolToPDBBlock(mol)
    elif extension in ['mol', 'mol2', 'sdf']:
        content = Chem.MolToMolBlock(mol, kekulize=False)
        extension = 'mol'
    else:
        raise NotImplementedError

    return content, extension, None


def show_target(in_protein):
    molecule, extension, html = read_molecule_file(in_protein, allowed_extentions=['pdb'])
    if molecule is not None:
        html = static.TARGET_RENDERING_TEMPLATE.format(molecule=molecule, fmt=extension)

    return static.IFRAME_TEMPLATE.format(html=html)

def show_complex(complex_path):
    protein_complex, extension, html = read_molecule_file(complex_path, allowed_extentions=['pdb'])
    if protein_complex is not None:
        html = static.COMPLEX_RENDERING_TEMPLATE.format(complex=protein_complex, fmt=extension)

    return static.IFRAME_TEMPLATE.format(html=html)

# def create_complex_view_html(
#         complex_path, pocket_path_dict=None,
#         interactive_ligands=True, interactive_pockets=True
# ):
#     """Generates HTML for complex visualization."""
#     model_i = -1
#     viewer_models = ""
#     if complex_path:
#         complex_data, extension, html = read_molecule_file(complex_path, allowed_extentions=['pdb'])
#         viewer_models += f'viewer.addModel(`{complex_data}`, "pdb");'
#         model_i += 1
#         viewer_models += f"viewer.getModel({model_i}).setStyle({{ hetflag: false }}, proteinStyle);"
#         viewer_models += f"viewer.getModel({model_i}).setStyle({{ hetflag: true }}, ligandStyle);"
#         if interactive_ligands:
#             # return ligand residue info when the ligand is clicked
#             viewer_models += f"""
#                 let selectedLigand = null;
#                 viewer.getModel({model_i}).setClickable(
#                     {{ hetflag: true, byres: true }},
#                     true,
#                     function (_atom, _viewer, _event, _container) {{
#                         let currentLigand = {{ resn: _atom.resn, chain: _atom.chain, resi: _atom.resi }};
#
#                         if (selectedLigand === currentLigand) {{
#                             // Deselect ligand
#                             selectedLigand = null;
#                             _viewer.setStyle(
#                                 {{ resn: _atom.resn, chain: _atom.chain, resi: _atom.resi }},
#                                 ligandStyle
#                             );
#                             console.log("Deselected Residue:", currentLigand);
#                             window.parent.postMessage({{
#                                 name: "ligand_selection",
#                                 data: {{ residue: currentLigand, add: false }}
#                             }}, "*");
#                         }} else {{
#                             // Select ligand and deselect previous
#                             if (selectedLigand) {{
#                                 _viewer.setStyle(
#                                     {{
#                                         resn: selectedLigand.resn,
#                                         chain: selectedLigand.chain,
#                                         resi: selectedLigand.resi
#                                     }},
#                                     ligandStyle
#                                 );
#                             }}
#                             selectedLigand = currentLigand;
#                             _viewer.setStyle(
#                                 {{ resn: _atom.resn, chain: _atom.chain, resi: _atom.resi }},
#                                 {{ stick: {{ color: "red", radius: 0.4}} }}
#                             );
#                             console.log("Selected Residue:", currentLigand);
#                             window.parent.postMessage({{
#                                 name: "ligand_selection",
#                                 data: {{ residue: currentLigand, add: true }}
#                             }}, "*");
#                         }}
#                         _viewer.render();
#                     }}
#                 );
#             """
#     if pocket_path_dict:
#         pocket_data_dict = {k: open(v, 'r').read() for k, v in pocket_path_dict.items()}
#         for pocket_name, pocket_data in pocket_data_dict.items():
#             viewer_models += f'viewer.addModel(`{pocket_data}`, "pqr");'
#             model_i += 1
#             viewer_models += f'viewer.getModel({model_i}).setStyle(pocketStyle);'
#             if interactive_pockets:
#                 # return the pocket name when the pocket is clicked
#                 viewer_models += f"""
#                     let selectedPocket = null;
#                     viewer.getModel({model_i}).setClickable(
#                         {{ byres: true }},
#                         true,
#                         function (_atom, _viewer, _event, _container) {{
#                             let currentPocket = "{pocket_name}";
#
#                             if (selectedPocket == currentPocket) {{
#                                 // Deselect pocket
#                                 selectedPocket = null;
#                                 _viewer.getModel({model_i}).setStyle( pocketStyle );
#                                 console.log("Deselected Pocket:", currentPocket);
#                                 window.parent.postMessage({{
#                                     name: "pocket_selection",
#                                     data: {{ pocket: currentPocket, add: false }}
#                                 }}, "*");
#                             }} else {{
#                                 // Select pocket and deselect previous
#                                 if (selectedPocket) {{
#                                     _viewer.getModel(selectedPocket).setStyle( pocketStyle );
#                                 }}
#                                 selectedPocket = currentPocket;
#                                 _viewer.getModel({model_i}).setStyle(
#                                     {{ sphere: {{ color: "red", opacity: 0.9}} }}
#                                 );
#                                 console.log("Selected Pocket:", currentPocket);
#                                 window.parent.postMessage({{
#                                     name: "pocket_selection",
#                                     data: {{ pocket: currentPocket, add: true }}
#                                 }}, "*");
#                             }}
#                             _viewer.render();
#                         }}
#                     );
#                 """
#
#     html = static.COMPLEX_RENDERING_TEMPLATE.format(viewer_models=viewer_models)
#     return static.IFRAME_TEMPLATE.format(html=html)

def create_result_table_html(summary_df, opts=(), progress=gr.Progress(track_tqdm=True)):
    html_df = summary_df.copy()
    column_aliases = {
        'ID1': 'Compound ID',
        'ID2': 'Target ID',
        'X1': 'Compound SMILES',
        'ligand_conf_path': 'Pose',
        'output_path': 'Pose'
    }
    # drop any columns ending with '_path'
    hidden_cols = [col for col in html_df.columns if col.endswith('_path')]

    html_df.rename(columns=column_aliases, inplace=True)
    if 'Compound' in html_df.columns and 'Exclude Molecular Graph' not in opts:
        html_df['Compound'] = html_df['Compound'].apply(PandasTools.PrintAsImageString)
    else:
        html_df.drop(['Compound'], axis=1, inplace=True)

    # if 'Scaffold' in html_df.columns and 'Exclude Scaffold Graph' not in opts:
    #     html_df['Scaffold'] = html_df['Scaffold'].parallel_apply(
    #         lambda x: PandasTools.PrintAsImageString(x) if not pd.isna(x) else x)
    # else:
    #     html_df.drop(['Scaffold'], axis=1, inplace=True)

    # html_df.index.name = 'Index'

    num_cols = html_df.select_dtypes('number').columns
    num_col_colors = sns.color_palette('husl', len(num_cols))
    bool_cols = html_df.select_dtypes(bool).columns

    image_zoom_formatter = HTMLTemplateFormatter(template='<div class="image-zoom-viewer"><%= value %></div>')
    uniprot_id_formatter = HTMLTemplateFormatter(
        template='<% if (value == value) { '  # Check if value is not NaN
                 'if (/^[OPQ][0-9][A-Z0-9]{3}[0-9]|[A-NR-Z][0-9]([A-Z][A-Z0-9]{2}[0-9]){1,2}$/.test(value)) '
        # Check if value is a valid UniProt ID
                 '{ %><a href="https://www.uniprot.org/uniprotkb/<%= value %>" target="_blank"><%= value %></a><% '
        # Else treat it as a sequence or other plain-text string, line-warping every 60 characters
                 '} else { %><div style="white-space: pre-wrap;"><%= value.match(/.{1,60}/g).join("<br>") '
                 '%></div><% } %><% } else { %><% } %>'  # Output empty string if value is NaN
    )
    pubchem_id_formatter = HTMLTemplateFormatter(
        template='<% if (value == value) { '  # Check if value is not NaN
                 '%><a href="https://pubchem.ncbi.nlm.nih.gov/#query=<%= value %>" '
                 'target="_blank"><%= value %></a>'
                 '<% } else { %><% } %>'  # Output empty string if value is NaN
    )
    bool_formatters = {col: BooleanFormatter() for col in bool_cols}
    float_formatters = {col: NumberFormatter(format='0.000') for col in html_df.select_dtypes('floating').columns}
    other_formatters = {
        'Compound': image_zoom_formatter,
        # 'Scaffold': image_zoom_formatter,
        # 'Target FASTA': {'type': 'textarea', 'width': 60},
        'Target ID': uniprot_id_formatter,
        # 'Compound ID': pubchem_id_formatter,  ## TODO: add link to click for adding mol to the viewer
        'Pose': {'type': 'molDisplayButtonFormatter'},
    }
    formatters = {**bool_formatters, **float_formatters, **other_formatters}

    # html = df.to_html(file)
    # return html

    report_table = pn.widgets.Tabulator(
        html_df, formatters=formatters,
        frozen_columns=['Compound ID', 'Compound'],
        hidden_columns=hidden_cols,
        disabled=True, sizing_mode='stretch_both', pagination='local', page_size=10
    )

    for i, col in enumerate(num_cols):
        cmap = sns.light_palette(num_col_colors[i], as_cmap=True)
        cmap.set_bad(color='white')
        report_table.style.background_gradient(
            subset=html_df.columns == col, cmap=cmap)

    # TODO change this to use commonn substructures
    pie_charts = {}
    # for y in html_df.columns.intersection(['Interaction Probability', 'Binding Affinity (IC50 [nM])']):
    #     for category in categories:
    #         pie_charts[y][category] = []
    #         for k in [10, 30, 100]:
    #             if k < len(html_df):
    #                 pie_charts[y][category].append(create_pie_chart(html_df, category=category, value=y, top_k=k))
    #             else:
    #                 pie_charts[y][category].append(create_pie_chart(html_df, category=category, value=y, top_k=len(html_df)))
    #                 break
    #         # Add 'All' tab regardless of the prediction dataset size
    #         # pie_charts[y].append(create_pie_chart(html_df, category=category, value=y, top_k=len(html_df)))
    #
    #         # Remove key-value pairs with an empty list
    #         pie_charts[y] = {k: v for k, v in pie_charts[y].items() if any(v)}
    #     pie_charts = {k: v for k, v in pie_charts.items() if any(v)}

    panel_css = """
        .tabulator {
            font-family: Courier New !important;
            font-weight: normal !important;
            font-size: 12px !important;
        }
        
        .tabulator-cell {
            overflow: visible !important;
            align-content: center !important;
        }
        
        .tabulator-cell:hover {
            z-index: 1000 !important;
        }
        
        .image-zoom-viewer {
            display: inline-block;
            overflow: visible;
            z-index: 1000;
        }
        
        .image-zoom-viewer::after {
            content: "";
            top: 0;
            left: 0;
            width: 100%;
            height: 100%;
            pointer-events: none;
        }
        
        .image-zoom-viewer:hover::after {
            pointer-events: all;
        }
        
        /* When hovering over the container, scale its child (the SVG) */
        .tabulator-cell:hover .image-zoom-viewer svg {
            padding: 3px;
            position: absolute;
            background-color: rgba(250, 250, 250, 0.854);
            box-shadow: 0 0 10px rgba(0, 0, 0, 0.618);
            border-radius: 3px;
            transform: scale(3); /* Scale up the SVG */
            transition: transform 0.3s ease;
            pointer-events: none; /* Prevents the SVG from blocking mouse interactions */
            z-index: 1000;
        }
    """

    pn.extension(
        raw_css=[panel_css],
        js_files={'panel_custom': 'app/panel.js'},
        # js_modules={'3Dmol': 'static/3Dmol-min.js'},
        inline=True,
    )

    template = pn.template.VanillaTemplate(
        sidebar=[],
        header=False,
        busy_indicator=None,
    )

    # stats_pane = pn.Column()
    # if pie_charts:
    #     for score_name, figure_dict in pie_charts.items():
    #         score_row = pn.Row()
    #         for category, figure_list in figure_dict.items():
    #             score_row.append(
    #                 pn.Column(f'### {category} by Top {score_name}', pn.Tabs(*figure_list, tabs_location='above')),
    #                 # pn.Card(pn.Row(v), title=f'{category} by Top {k}')
    #             )
    #         stats_pane.append(
    #             score_row
    #         )
    #
    # if stats_pane:
    #     template.main.append(
    #         pn.Card(stats_pane, sizing_mode='stretch_width', title='Summary Statistics', margin=10)
    #     )

    template.main.append(
        pn.Card(report_table, title=f'GenFBDD Results',  # width=1200,
                margin=10)
    )

    with tempfile.TemporaryDirectory() as tmpdir:
        file = Path(tmpdir) / 'report.html'
        template.save(file)
        html_str = file.read_text()
        iframe_html = static.IFRAME_TEMPLATE.format(html=html_str)
        return iframe_html


def pdb_query(query, method):
    """Downloads protein structure data or searches FASTA sequence."""
    gr.Info(f'Querying protein by {method}...')
    try:
        if method == 'PDB ID':
            url = f"https://files.rcsb.org/download/{query}.pdb"
            return download_file(url)
        elif method == 'UniProt ID':
            pdb_ids = uniprot_to_pdb(query)
            if pdb_ids:
                # Download the first associated PDB file
                return download_file(f"https://files.rcsb.org/download/{pdb_ids[0]}.pdb")
            else:
                raise ValueError(f"No PDB IDs found for UniProt ID: {query}")
        elif method == 'FASTA Sequence':
            pdb_ids = fasta_to_pdb(query)
            if pdb_ids:
                # Download the first associated PDB file
                return download_file(f"https://files.rcsb.org/download/{pdb_ids[0]}.pdb")
            else:
                raise ValueError("No PDB IDs found for the provided FASTA sequence.")
        else:
            raise ValueError(f"Unsupported method: {method}")
    except Exception as e:
        gr.Warning(f"Error downloading PDB file: {e}")
        return None


def download_file(url):
    """Downloads a small file to a temporary location, preserving its filename."""
    try:
        response = requests.get(url)
        response.raise_for_status()

        filename = Path(url).name
        temp_dir = Path(tempfile.gettempdir())
        temp_path = temp_dir / filename

        temp_path.write_bytes(response.content)

        return str(temp_path)
    except Exception as e:
        gr.Error(f"Download error: {e}")
        return None


def uniprot_to_pdb(uniprot_id):
    """Queries the RCSB PDB API to find PDB entities associated with a UniProt ID."""
    base_url = "https://search.rcsb.org/rcsbsearch/v2/query"
    query_payload = {
        "query": {
            "type": "group",
            "logical_operator": "and",
            "nodes": [
                {
                    "type": "terminal",
                    "service": "text",
                    "parameters": {
                        "operator": "exact_match",
                        "value": uniprot_id,
                        "attribute": "rcsb_polymer_entity_container_identifiers.reference_sequence_identifiers.database_accession"
                    }
                },
                {
                    "type": "terminal",
                    "service": "text",
                    "parameters": {
                        "operator": "exact_match",
                        "value": "UniProt",
                        "attribute": "rcsb_polymer_entity_container_identifiers.reference_sequence_identifiers.database_name"
                    }
                }
            ]
        },
        "return_type": "entry"
    }
    try:
        # Send POST request with JSON payload
        response = requests.post(base_url, json=query_payload)
        response.raise_for_status()
        data = response.json()
        return [entry["identifier"] for entry in data.get("result_set", [])]
    except Exception as e:
        print(f"Error querying UniProt ID: {e}")
        return []


def fasta_to_pdb(fasta_sequence):
    """Queries the RCSB PDB API to find PDB IDs associated with a FASTA sequence."""
    base_url = "https://search.rcsb.org/rcsbsearch/v2/query"
    query_payload = {
        "query": {
            "type": "terminal",
            "service": "sequence",
            "parameters": {
                "evalue_cutoff": 1,
                "identity_cutoff": 0.9,
                "sequence_type": "protein",
                "value": fasta_sequence
            }
        },
        "request_options": {
            "scoring_strategy": "sequence"
        },
        "return_type": "entry"
    }
    try:
        # Send POST request with JSON payload
        response = requests.post(base_url, json=query_payload)
        response.raise_for_status()
        data = response.json()
        return [entry["identifier"] for entry in data.get("result_set", [])]
    except Exception as e:
        print(f"Error querying FASTA sequence: {e}")
        return []