app.py

import copy
import re

import os.path

import torch
import sys

import gradio as gr
import pandas as pd
import numpy as np
import plotly.express as px
import matplotlib.pyplot as plt
import plotly.graph_objects as go

import tempfile
import requests
from moleculekit.molecule import Molecule

sys.path.append("/home/user/app/ProteinMPNN/vanilla_proteinmpnn")
# this is for local
sys.path.append(os.path.join(os.getcwd(), "ProteinMPNN/vanilla_proteinmpnn"))


def make_tied_positions_for_homomers(pdb_dict_list):
    my_dict = {}
    for result in pdb_dict_list:
        all_chain_list = sorted(
            [item[-1:] for item in list(result) if item[:9] == "seq_chain"]
        )  # A, B, C, ...
        tied_positions_list = []
        chain_length = len(result[f"seq_chain_{all_chain_list[0]}"])
        for i in range(1, chain_length + 1):
            temp_dict = {}
            for j, chain in enumerate(all_chain_list):
                temp_dict[chain] = [i]  # needs to be a list
            tied_positions_list.append(temp_dict)
        my_dict[result["name"]] = tied_positions_list
    return my_dict


def align_structures(pdb1, pdb2, index):
    """Take two structure and superimpose pdb1 on pdb2"""
    import Bio.PDB
    import subprocess

    pdb_parser = Bio.PDB.PDBParser(QUIET=True)
    # Get the structures
    ref_structure = pdb_parser.get_structure("ref", pdb1)
    sample_structure = pdb_parser.get_structure("sample", pdb2)

    sample_structure_ca = [
        atom for atom in sample_structure.get_atoms() if atom.name == "CA"
    ]
    plddts = [atom.get_bfactor() for atom in sample_structure_ca]

    aligner = Bio.PDB.CEAligner()
    aligner.set_reference(ref_structure)
    aligner.align(sample_structure)

    io = Bio.PDB.PDBIO()
    io.set_structure(ref_structure)
    hash = os.path.splitext(os.path.basename(pdb2))[0]
    io.save(f"outputs/{hash}_ref_{index}.pdb")
    io.set_structure(sample_structure)
    io.save(f"outputs/{hash}_align_{index}.pdb")
    # Doing this to get around biopython CEALIGN bug
    # subprocess.call("pymol -c -Q -r cealign.pml", shell=True)

    return (
        aligner.rms,
        f"outputs/{hash}_ref_{index}.pdb",
        f"outputs/{hash}_align_{index}.pdb",
        plddts,
    )


if not os.path.exists("/home/user/app/ProteinMPNN/"):
    path_to_model_weights = os.path.join(
        os.getcwd(), "ProteinMPNN/vanilla_proteinmpnn/vanilla_model_weights"
    )
    is_local = True
else:
    path_to_model_weights = (
        "/home/user/app/ProteinMPNN/vanilla_proteinmpnn/vanilla_model_weights"
    )
    is_local = False


if is_local:
    print("Running locally")
    from transformers import AutoTokenizer, EsmForProteinFolding


def setup_proteinmpnn(model_name="v_48_020", backbone_noise=0.00):
    from protein_mpnn_utils import (
        loss_nll,
        loss_smoothed,
        gather_edges,
        gather_nodes,
        gather_nodes_t,
        cat_neighbors_nodes,
        _scores,
        _S_to_seq,
        tied_featurize,
        parse_PDB,
    )
    from protein_mpnn_utils import StructureDataset, StructureDatasetPDB, ProteinMPNN

    device = torch.device(
        "cpu"
    )  # torch.device("cuda:0" if (torch.cuda.is_available()) else "cpu") #fix for memory issues
    # ProteinMPNN model name: v_48_002, v_48_010, v_48_020, v_48_030, v_32_002, v_32_010; v_32_020, v_32_030; v_48_010=version with 48 edges 0.10A noise
    # Standard deviation of Gaussian noise to add to backbone atoms
    hidden_dim = 128
    num_layers = 3
    model_folder_path = path_to_model_weights
    if model_folder_path[-1] != "/":
        model_folder_path = model_folder_path + "/"
    checkpoint_path = model_folder_path + f"{model_name}.pt"

    checkpoint = torch.load(checkpoint_path, map_location=device)

    noise_level_print = checkpoint["noise_level"]

    model = ProteinMPNN(
        num_letters=21,
        node_features=hidden_dim,
        edge_features=hidden_dim,
        hidden_dim=hidden_dim,
        num_encoder_layers=num_layers,
        num_decoder_layers=num_layers,
        augment_eps=backbone_noise,
        k_neighbors=checkpoint["num_edges"],
    )
    model.to(device)
    model.load_state_dict(checkpoint["model_state_dict"])
    model.eval()
    return model, device


def get_pdb(pdb_code="", filepath=""):
    if pdb_code is None or pdb_code == "":
        try:
            return filepath.name
        except AttributeError as e:
            return None
    else:
        os.system(f"wget -qnc https://files.rcsb.org/view/{pdb_code}.pdb")
        return f"{pdb_code}.pdb"


def preprocess_mol(pdb_code="", filepath=""):
    if pdb_code is None or pdb_code == "":
        try:
            mol = Molecule(filepath.name)
        except AttributeError as e:
            return None
    else:
        mol = Molecule(pdb_code)
    mol.write("original.pdb")
    # clean messy files and only include protein itself
    mol.filter("protein")
    # renumber using moleculekit 0...len(protein)
    df = mol.renumberResidues(returnMapping=True)
    # add proteinMPNN index col which used 1..len(chain), 1...len(chain)
    indexes = []
    for chain, g in df.groupby("chain"):
        j = 1
        for i, row in g.iterrows():
            indexes.append(j)
            j += 1
    df["proteinMPNN_index"] = indexes
    mol.write("cleaned.pdb")
    return "cleaned.pdb", df


def assign_sasa(mol):
    from moleculekit.projections.metricsasa import MetricSasa

    metr = MetricSasa(mode="residue", filtersel="protein")
    sasaR = metr.project(mol)[0]
    is_prot = mol.atomselect("protein")
    resids = pd.DataFrame.from_dict({"resid": mol.resid, "is_prot": is_prot})
    new_masses = []
    i_without_non_prot = 0
    for i, g in resids.groupby((resids["resid"].shift() != resids["resid"]).cumsum()):
        if g["is_prot"].unique()[0] == True:
            g["sasa"] = sasaR[i_without_non_prot]
            i_without_non_prot += 1
        else:
            g["sasa"] = 0
        new_masses.extend(list(g.sasa))
    return np.array(new_masses)


def process_atomsel(atomsel):
    """everything lowercase and replace some keywords not relevant for protein design"""
    atomsel = re.sub("sasa", "mass", atomsel, flags=re.I)
    atomsel = re.sub("plddt", "beta", atomsel, flags=re.I)
    return atomsel


def make_fixed_positions_dict(atomsel, residue_index_df):
    # we use the uploaded file for the selection
    mol = Molecule("original.pdb")
    # use index for selection as resids will change

    # set sasa to 0 for all non protein atoms (all non protein atoms are deleted later)
    mol.masses = assign_sasa(mol)
    print(mol.masses.shape)
    print(assign_sasa(mol).shape)
    atomsel = process_atomsel(atomsel)
    selected_residues = mol.get("index", atomsel)

    # clean up
    mol.filter("protein")
    mol.renumberResidues()
    # based on selected index now get resids
    selected_residues = [str(i) for i in selected_residues]
    if len(selected_residues) == 0:
        return None, []
    selected_residues_str = " ".join(selected_residues)
    selected_residues = set(mol.get("resid", sel=f"index {selected_residues_str}"))

    # use the proteinMPNN index nomenclature to assemble fixed_positions_dict
    fixed_positions_df = residue_index_df[
        residue_index_df["new_resid"].isin(selected_residues)
    ]

    chains = set(mol.get("chain", sel="all"))
    fixed_position_dict = {"cleaned": {}}
    # store the selected residues in a list for the visualization later with cleaned.pdb
    selected_residues = list(fixed_positions_df["new_resid"])

    for c in chains:
        fixed_position_dict["cleaned"][c] = []

    for i, row in fixed_positions_df.iterrows():
        fixed_position_dict["cleaned"][row["chain"]].append(row["proteinMPNN_index"])
    return fixed_position_dict, selected_residues


def update(
    inp,
    file,
    designed_chain,
    fixed_chain,
    homomer,
    num_seqs,
    sampling_temp,
    model_name,
    backbone_noise,
    atomsel,
):
    from protein_mpnn_utils import (
        loss_nll,
        loss_smoothed,
        gather_edges,
        gather_nodes,
        gather_nodes_t,
        cat_neighbors_nodes,
        _scores,
        _S_to_seq,
        tied_featurize,
        parse_PDB,
    )
    from protein_mpnn_utils import StructureDataset, StructureDatasetPDB, ProteinMPNN

    # pdb_path = get_pdb(pdb_code=inp, filepath=file)

    pdb_path, mol_index = preprocess_mol(pdb_code=inp, filepath=file)

    if pdb_path == None:
        return "Error processing PDB"

    model, device = setup_proteinmpnn(
        model_name=model_name, backbone_noise=backbone_noise
    )

    if designed_chain == "":
        designed_chain_list = []
    else:
        designed_chain_list = re.sub("[^A-Za-z]+", ",", designed_chain).split(",")

    if fixed_chain == "":
        fixed_chain_list = []
    else:
        fixed_chain_list = re.sub("[^A-Za-z]+", ",", fixed_chain).split(",")

    chain_list = list(set(designed_chain_list + fixed_chain_list))
    num_seq_per_target = num_seqs
    save_score = 0  # 0 for False, 1 for True; save score=-log_prob to npy files
    save_probs = (
        0  # 0 for False, 1 for True; save MPNN predicted probabilites per position
    )
    score_only = 0  # 0 for False, 1 for True; score input backbone-sequence pairs
    conditional_probs_only = 0  # 0 for False, 1 for True; output conditional probabilities p(s_i given the rest of the sequence and backbone)
    conditional_probs_only_backbone = 0  # 0 for False, 1 for True; if true output conditional probabilities p(s_i given backbone)

    batch_size = 1  # Batch size; can set higher for titan, quadro GPUs, reduce this if running out of GPU memory
    max_length = 20000  # Max sequence length

    out_folder = "."  # Path to a folder to output sequences, e.g. /home/out/
    jsonl_path = ""  # Path to a folder with parsed pdb into jsonl
    omit_AAs = "X"  # Specify which amino acids should be omitted in the generated sequence, e.g. 'AC' would omit alanine and cystine.

    pssm_multi = 0.0  # A value between [0.0, 1.0], 0.0 means do not use pssm, 1.0 ignore MPNN predictions
    pssm_threshold = 0.0  # A value between -inf + inf to restric per position AAs
    pssm_log_odds_flag = 0  # 0 for False, 1 for True
    pssm_bias_flag = 0  # 0 for False, 1 for True

    folder_for_outputs = out_folder

    NUM_BATCHES = num_seq_per_target // batch_size
    BATCH_COPIES = batch_size
    temperatures = [sampling_temp]
    omit_AAs_list = omit_AAs
    alphabet = "ACDEFGHIKLMNPQRSTVWYX"

    omit_AAs_np = np.array([AA in omit_AAs_list for AA in alphabet]).astype(np.float32)

    chain_id_dict = None
    if atomsel == "":
        fixed_positions_dict, selected_residues = None, []
    else:
        fixed_positions_dict, selected_residues = make_fixed_positions_dict(
            atomsel, mol_index
        )

    pssm_dict = None
    omit_AA_dict = None
    bias_AA_dict = None

    bias_by_res_dict = None
    bias_AAs_np = np.zeros(len(alphabet))

    ###############################################################
    pdb_dict_list = parse_PDB(pdb_path, input_chain_list=chain_list)
    dataset_valid = StructureDatasetPDB(
        pdb_dict_list, truncate=None, max_length=max_length
    )
    if homomer:
        tied_positions_dict = make_tied_positions_for_homomers(pdb_dict_list)
    else:
        tied_positions_dict = None

    chain_id_dict = {}
    chain_id_dict[pdb_dict_list[0]["name"]] = (designed_chain_list, fixed_chain_list)
    with torch.no_grad():
        for ix, prot in enumerate(dataset_valid):
            score_list = []
            all_probs_list = []
            all_log_probs_list = []
            S_sample_list = []
            batch_clones = [copy.deepcopy(prot) for i in range(BATCH_COPIES)]
            (
                X,
                S,
                mask,
                lengths,
                chain_M,
                chain_encoding_all,
                chain_list_list,
                visible_list_list,
                masked_list_list,
                masked_chain_length_list_list,
                chain_M_pos,
                omit_AA_mask,
                residue_idx,
                dihedral_mask,
                tied_pos_list_of_lists_list,
                pssm_coef,
                pssm_bias,
                pssm_log_odds_all,
                bias_by_res_all,
                tied_beta,
            ) = tied_featurize(
                batch_clones,
                device,
                chain_id_dict,
                fixed_positions_dict,
                omit_AA_dict,
                tied_positions_dict,
                pssm_dict,
                bias_by_res_dict,
            )
            pssm_log_odds_mask = (
                pssm_log_odds_all > pssm_threshold
            ).float()  # 1.0 for true, 0.0 for false
            name_ = batch_clones[0]["name"]

            randn_1 = torch.randn(chain_M.shape, device=X.device)
            log_probs = model(
                X,
                S,
                mask,
                chain_M * chain_M_pos,
                residue_idx,
                chain_encoding_all,
                randn_1,
            )
            mask_for_loss = mask * chain_M * chain_M_pos
            scores = _scores(S, log_probs, mask_for_loss)
            native_score = scores.cpu().data.numpy()
            message = ""
            seq_list = []
            seq_recovery = []
            seq_score = []
            for temp in temperatures:
                for j in range(NUM_BATCHES):
                    randn_2 = torch.randn(chain_M.shape, device=X.device)
                    if tied_positions_dict == None:
                        sample_dict = model.sample(
                            X,
                            randn_2,
                            S,
                            chain_M,
                            chain_encoding_all,
                            residue_idx,
                            mask=mask,
                            temperature=temp,
                            omit_AAs_np=omit_AAs_np,
                            bias_AAs_np=bias_AAs_np,
                            chain_M_pos=chain_M_pos,
                            omit_AA_mask=omit_AA_mask,
                            pssm_coef=pssm_coef,
                            pssm_bias=pssm_bias,
                            pssm_multi=pssm_multi,
                            pssm_log_odds_flag=bool(pssm_log_odds_flag),
                            pssm_log_odds_mask=pssm_log_odds_mask,
                            pssm_bias_flag=bool(pssm_bias_flag),
                            bias_by_res=bias_by_res_all,
                        )
                        S_sample = sample_dict["S"]
                    else:
                        sample_dict = model.tied_sample(
                            X,
                            randn_2,
                            S,
                            chain_M,
                            chain_encoding_all,
                            residue_idx,
                            mask=mask,
                            temperature=temp,
                            omit_AAs_np=omit_AAs_np,
                            bias_AAs_np=bias_AAs_np,
                            chain_M_pos=chain_M_pos,
                            omit_AA_mask=omit_AA_mask,
                            pssm_coef=pssm_coef,
                            pssm_bias=pssm_bias,
                            pssm_multi=pssm_multi,
                            pssm_log_odds_flag=bool(pssm_log_odds_flag),
                            pssm_log_odds_mask=pssm_log_odds_mask,
                            pssm_bias_flag=bool(pssm_bias_flag),
                            tied_pos=tied_pos_list_of_lists_list[0],
                            tied_beta=tied_beta,
                            bias_by_res=bias_by_res_all,
                        )
                        # Compute scores
                        S_sample = sample_dict["S"]
                    log_probs = model(
                        X,
                        S_sample,
                        mask,
                        chain_M * chain_M_pos,
                        residue_idx,
                        chain_encoding_all,
                        randn_2,
                        use_input_decoding_order=True,
                        decoding_order=sample_dict["decoding_order"],
                    )
                    mask_for_loss = mask * chain_M * chain_M_pos
                    scores = _scores(S_sample, log_probs, mask_for_loss)
                    scores = scores.cpu().data.numpy()
                    all_probs_list.append(sample_dict["probs"].cpu().data.numpy())
                    all_log_probs_list.append(log_probs.cpu().data.numpy())
                    S_sample_list.append(S_sample.cpu().data.numpy())
                    for b_ix in range(BATCH_COPIES):
                        masked_chain_length_list = masked_chain_length_list_list[b_ix]
                        masked_list = masked_list_list[b_ix]
                        seq_recovery_rate = torch.sum(
                            torch.sum(
                                torch.nn.functional.one_hot(S[b_ix], 21)
                                * torch.nn.functional.one_hot(S_sample[b_ix], 21),
                                axis=-1,
                            )
                            * mask_for_loss[b_ix]
                        ) / torch.sum(mask_for_loss[b_ix])
                        seq = _S_to_seq(S_sample[b_ix], chain_M[b_ix])
                        score = scores[b_ix]
                        score_list.append(score)
                        native_seq = _S_to_seq(S[b_ix], chain_M[b_ix])
                        if b_ix == 0 and j == 0 and temp == temperatures[0]:
                            start = 0
                            end = 0
                            list_of_AAs = []
                            for mask_l in masked_chain_length_list:
                                end += mask_l
                                list_of_AAs.append(native_seq[start:end])
                                start = end
                            native_seq = "".join(
                                list(np.array(list_of_AAs)[np.argsort(masked_list)])
                            )
                            l0 = 0
                            for mc_length in list(
                                np.array(masked_chain_length_list)[
                                    np.argsort(masked_list)
                                ]
                            )[:-1]:
                                l0 += mc_length
                                native_seq = native_seq[:l0] + "/" + native_seq[l0:]
                                l0 += 1
                            sorted_masked_chain_letters = np.argsort(
                                masked_list_list[0]
                            )
                            print_masked_chains = [
                                masked_list_list[0][i]
                                for i in sorted_masked_chain_letters
                            ]
                            sorted_visible_chain_letters = np.argsort(
                                visible_list_list[0]
                            )
                            print_visible_chains = [
                                visible_list_list[0][i]
                                for i in sorted_visible_chain_letters
                            ]
                            native_score_print = np.format_float_positional(
                                np.float32(native_score.mean()),
                                unique=False,
                                precision=4,
                            )
                            line = ">{}, score={}, fixed_chains={}, designed_chains={}, model_name={}\n{}\n".format(
                                name_,
                                native_score_print,
                                print_visible_chains,
                                print_masked_chains,
                                model_name,
                                native_seq,
                            )
                            message += f"{line}\n"
                        start = 0
                        end = 0
                        list_of_AAs = []
                        for mask_l in masked_chain_length_list:
                            end += mask_l
                            list_of_AAs.append(seq[start:end])
                            start = end

                        seq = "".join(
                            list(np.array(list_of_AAs)[np.argsort(masked_list)])
                        )
                        # add non designed chains to predicted sequence
                        l0 = 0
                        for mc_length in list(
                            np.array(masked_chain_length_list)[np.argsort(masked_list)]
                        )[:-1]:
                            l0 += mc_length
                            seq = seq[:l0] + "/" + seq[l0:]
                            l0 += 1
                        score_print = np.format_float_positional(
                            np.float32(score), unique=False, precision=4
                        )
                        seq_rec_print = np.format_float_positional(
                            np.float32(seq_recovery_rate.detach().cpu().numpy()),
                            unique=False,
                            precision=4,
                        )
                        chain_s = ""
                        if len(visible_list_list[0]) > 0:
                            chain_M_bool = chain_M.bool()
                            not_designed = _S_to_seq(S[b_ix], ~chain_M_bool[b_ix])

                            labels = (
                                chain_encoding_all[b_ix][~chain_M_bool[b_ix]]
                                .detach()
                                .cpu()
                                .numpy()
                            )

                            for c in set(labels):
                                chain_s += ":"
                                nd_mask = labels == c
                                for i, x in enumerate(not_designed):
                                    if nd_mask[i]:
                                        chain_s += x
                        seq_recovery.append(seq_rec_print)
                        seq_score.append(score_print)
                        line = (
                            ">T={}, sample={}, score={}, seq_recovery={}\n{}\n".format(
                                temp, b_ix, score_print, seq_rec_print, seq
                            )
                        )
                        seq_list.append(seq + chain_s)
                        message += f"{line}\n"
    if fixed_positions_dict != None:
        message += f"\nfixed positions:* {fixed_positions_dict['cleaned']} \n\n*uses CHAIN:[1..len(chain)] residue numbering"
    # somehow sequences still contain X, remove again
    for i, x in enumerate(seq_list):
        for aa in omit_AAs:
            seq_list[i] = x.replace(aa, "")
    all_probs_concat = np.concatenate(all_probs_list)
    all_log_probs_concat = np.concatenate(all_log_probs_list)
    np.savetxt("all_probs_concat.csv", all_probs_concat.mean(0).T, delimiter=",")
    np.savetxt(
        "all_log_probs_concat.csv",
        np.exp(all_log_probs_concat).mean(0).T,
        delimiter=",",
    )
    S_sample_concat = np.concatenate(S_sample_list)
    fig = px.imshow(
        np.exp(all_log_probs_concat).mean(0).T,
        labels=dict(x="positions", y="amino acids", color="probability"),
        y=list(alphabet),
        template="simple_white",
    )
    fig.update_xaxes(side="top")

    fig_tadjusted = px.imshow(
        all_probs_concat.mean(0).T,
        labels=dict(x="positions", y="amino acids", color="probability"),
        y=list(alphabet),
        template="simple_white",
    )

    fig_tadjusted.update_xaxes(side="top")
    seq_dict = {"seq_list": seq_list, "recovery": seq_recovery, "seq_score": seq_score}
    mol = structure_pred(seq_dict, pdb_path, selected_residues)
    print(seq_list)
    return (
        message,
        fig,
        fig_tadjusted,
        gr.File.update(value="all_log_probs_concat.csv", visible=True),
        gr.File.update(value="all_probs_concat.csv", visible=True),
        pdb_path,
        gr.Dropdown.update(choices=seq_list, value=seq_list[0], interactive=True),
        selected_residues,
        seq_dict,
        mol,
    )


def updateseq(seq, seq_dict, pdb_path, selected_residues):
    # find index of seq in seq_dict

    seq_list = seq_dict["seq_list"]
    seq_index = seq_list.index(seq)
    print(seq, seq_index)
    mol = structure_pred(seq_dict, pdb_path, selected_residues, index=seq_index)
    return mol


from transformers.models.esm.openfold_utils.protein import to_pdb, Protein as OFProtein
from transformers.models.esm.openfold_utils.feats import atom14_to_atom37


def convert_outputs_to_pdb(outputs):
    final_atom_positions = atom14_to_atom37(outputs["positions"][-1], outputs)
    outputs = {k: v.to("cpu").numpy() for k, v in outputs.items()}
    final_atom_positions = final_atom_positions.cpu().numpy()
    final_atom_mask = outputs["atom37_atom_exists"]
    pdbs = []
    for i in range(outputs["aatype"].shape[0]):
        aa = outputs["aatype"][i]
        pred_pos = final_atom_positions[i]
        mask = final_atom_mask[i]
        resid = outputs["residue_index"][i] + 1
        pred = OFProtein(
            aatype=aa,
            atom_positions=pred_pos,
            atom_mask=mask,
            residue_index=resid,
            b_factors=outputs["plddt"][i],
            chain_index=outputs["chain_index"][i] if "chain_index" in outputs else None,
        )
        pdbs.append(to_pdb(pred))
    return pdbs


def get_esmfold_local(sequence):

    filename = "outputs/" + hashlib.md5(str.encode(sequence)).hexdigest() + ".pdb"

    if not os.path.exists(filename):
        tokenizer = AutoTokenizer.from_pretrained("facebook/esmfold_v1")
        model = EsmForProteinFolding.from_pretrained(
            "facebook/esmfold_v1", low_cpu_mem_usage=True
        )

        model = model.cuda()
        model.esm = model.esm.half()
        import torch

        torch.backends.cuda.matmul.allow_tf32 = True
        model.trunk.set_chunk_size(64)

        position_id_offsets = []
        linker_mask = []
        for i, s in enumerate(sequence.split("/")):
            linker = 25 if i < sequence.count("/") else 0
            offsets = [i * 512] * (len(s) + linker)
            linker_mask.extend([1] * len(s) + [0] * linker)
            position_id_offsets.extend(offsets)
        sequence = sequence.replace("/", "G" * 25)
        tokenized = tokenizer([sequence], return_tensors="pt", add_special_tokens=False)
        with torch.no_grad():
            position_ids = torch.arange(len(sequence), dtype=torch.long)
            position_ids = position_ids + torch.torch.LongTensor(position_id_offsets)

        linker_mask = torch.Tensor(linker_mask).unsqueeze(1)
        tokenized["position_ids"] = position_ids.unsqueeze(0)
        tokenized = {key: tensor.cuda() for key, tensor in tokenized.items()}
        with torch.no_grad():
            output = model(**tokenized)

        output["atom37_atom_exists"] = output["atom37_atom_exists"] * linker_mask.to(
            output["atom37_atom_exists"].device
        )
        pdb = convert_outputs_to_pdb(output)

        with open(filename, "w+") as f:
            f.write("".join(pdb))
        print("local prediction", filename)
    else:
        print("prediction already on disk")
    return filename


def structure_pred(seq_dict, pdb, selectedResidues, index=0):
    allSeqs = seq_dict["seq_list"]
    lenSeqs = len(allSeqs)
    if len(allSeqs[index]) > 400:
        return """
        <div class="p-4 mb-4 text-sm text-yellow-700 bg-orange-50 rounded-lg" role="alert">
        <span class="font-medium">Sorry!</span> Currently only small proteins <400 aa can be predicted with the web api of ESMFold</div>
        """
    if "/" in allSeqs[index] and not is_local:
        return """
        <div class="p-4 mb-4 text-sm text-yellow-700 bg-orange-50 rounded-lg" role="alert">
        <span class="font-medium">Sorry!</span> Sequence is multimeric and no structure prediction is run. Use local copy of ESMFold to predict.</div>
        """
    i = 0
    sequences = {}
    if is_local:
        pdb_file = get_esmfold_local(allSeqs[index])
    else:
        pdb_file = get_esmfold(allSeqs[index])

    rms, input_pdb, aligned_pdb, plddts = align_structures(pdb, pdb_file, index)
    sequences[i] = {
        "Seq": index,
        "RMSD": f"{rms:.2f}",
        "Score": seq_dict["seq_score"][i],
        "Recovery": seq_dict["recovery"][i],
        "Mean pLDDT": f"{np.mean(plddts):.4f}",
    }

    num_res = len(allSeqs[index])
    return molecule(
        input_pdb,
        aligned_pdb,
        lenSeqs,
        num_res,
        selectedResidues,
        allSeqs,
        sequences,
    )


def read_mol(molpath):
    with open(molpath, "r") as fp:
        lines = fp.readlines()
    mol = ""
    for l in lines:
        mol += l
    return mol


def molecule(
    input_pdb, aligned_pdb, lenSeqs, num_res, selectedResidues, allSeqs, sequences
):
    print("mol updated")
    print("filenames", input_pdb, aligned_pdb)
    mol = read_mol(input_pdb)
    options = ""
    pred_mol = "["
    seqdata = "{"
    selected = "selected"
    for i in range(1):  # lenSeqs):
        seqdata += (
            str(sequences[i]["Seq"])
            + ': { "score": '
            + sequences[i]["Score"]
            + ', "rmsd": '
            + sequences[i]["RMSD"]
            + ', "recovery": '
            + sequences[i]["Recovery"]
            + ', "plddt": '
            + sequences[i]["Mean pLDDT"]
            + ', "seq":"'
            + allSeqs[i]
            + '"}'
        )

        pred_mol += f"`{read_mol(aligned_pdb)}`"
        selected = ""
        # if i != lenSeqs - 1:
        #     pred_mol += ","
        #     seqdata += ","
    pred_mol += "]"
    seqdata += "}"

    x = (
        """<!DOCTYPE html>
        <html>
        <head>    
    <meta http-equiv="content-type" content="text/html; charset=UTF-8" />
     <link rel="stylesheet" href="https://unpkg.com/[email protected]/dist/flowbite.min.css" />
    <style>
    body{
        font-family:sans-serif
    }
    .mol-container {
    width: 100%;
    height: 700px;
    position: relative;
    }
    .space-x-2 > * + *{
        margin-left: 0.5rem;
    }
    .p-1{
        padding:0.5rem;
    }
    .w-4{
        width:1rem;
    }
    .h-4{
        height:1rem;
    }
    .mt-4{
        margin-top:1rem;
    }
    .mol-container select{
        background-image:None;
    }
    </style>
    <script src="https://cdnjs.cloudflare.com/ajax/libs/jquery/3.6.3/jquery.min.js" integrity="sha512-STof4xm1wgkfm7heWqFJVn58Hm3EtS31XFaagaa8VMReCXAkQnJZ+jEy8PCC/iT18dFy95WcExNHFTqLyp72eQ==" crossorigin="anonymous" referrerpolicy="no-referrer"></script>

    <script src="https://3Dmol.csb.pitt.edu/build/3Dmol-min.js"></script>
    </head>
    <body>
    <div class="font-mono bg-gray-100 py-3 px-2  font-sm rounded">
        <code>> seq <span id="id"></span>, score <span id="score"></span>, RMSD <span id="seqrmsd"></span>, Recovery
            <span id="recovery"></span>, pLDDT <span id="plddt"></span></code><br>
        <p id="seqText" class="max-w-4xl font-xs block" style="word-break: break-all;">

        </p>
    </div>
    <div id="container" class="mol-container"></div>
    <div class="flex items-center">
        <div class="px-4 pt-2">
        <label for="sidechain" class="relative inline-flex items-center mb-4 cursor-pointer ">
            <input  id="sidechain" type="checkbox" class="sr-only peer">
            <div class="w-11 h-6 bg-gray-200 rounded-full peer peer-focus:ring-4 peer-focus:ring-blue-300 dark:peer-focus:ring-blue-800 dark:bg-gray-700 peer-checked:after:translate-x-full peer-checked:after:border-white after:absolute after:top-0.5 after:left-[2px] after:bg-white after:border-gray-300 after:border after:rounded-full after:h-5 after:w-5 after:transition-all dark:border-gray-600 peer-checked:bg-blue-600"></div>
            <span class="ml-3 text-sm font-medium text-gray-900 dark:text-gray-300">Show side chains</span>
          </label>
        </div>
        <div class="px-4 pt-2">
        <label for="startstructure" class="relative inline-flex items-center mb-4 cursor-pointer ">
            <input  id="startstructure" type="checkbox" class="sr-only peer" checked>
            <div class="w-11 h-6 bg-gray-200 rounded-full peer peer-focus:ring-4 peer-focus:ring-blue-300 dark:peer-focus:ring-blue-800 dark:bg-gray-700 peer-checked:after:translate-x-full peer-checked:after:border-white after:absolute after:top-0.5 after:left-[2px] after:bg-white after:border-gray-300 after:border after:rounded-full after:h-5 after:w-5 after:transition-all dark:border-gray-600 peer-checked:bg-blue-600"></div>
            <span class="ml-3 text-sm font-medium text-gray-900 dark:text-gray-300">Show input structure</span>
          </label>
        </div>
        <button type="button" class="text-gray-900 bg-white hover:bg-gray-100 border border-gray-200 focus:ring-4 focus:outline-none focus:ring-gray-100 font-medium rounded-lg text-sm px-5 py-2.5 text-center inline-flex items-center dark:focus:ring-gray-600 dark:bg-gray-800 dark:border-gray-700 dark:text-white dark:hover:bg-gray-700 mr-2 mb-2" id="download">
                    <svg class="w-6 h-6 mr-2 -ml-1" fill="none" stroke="currentColor" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg"><path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M4 16v1a3 3 0 003 3h10a3 3 0 003-3v-1m-4-4l-4 4m0 0l-4-4m4 4V4"></path></svg>
                    Download predicted structure
                  </button>
            </div>       
            <div class="text-sm">
            <div> RMSD ESMFold vs. native: <span id="rmsd"></span> Å computed using CEAlign on the aligned fragment</div>
                                    </div>
            <div class="text-sm flex items-start">
                <div class="w-1/2">
                        
                            <div class="font-medium mt-4 flex items-center space-x-2"><b>AF2 model of redesigned sequence</b></div>
                            <div>ESMFold model confidence:</div>
                            <div class="flex space-x-2 py-1"><span class="w-4 h-4" style="background-color: rgb(0, 83, 214);">&nbsp;</span><span class="legendlabel">Very high
                                    (pLDDT &gt; 90)</span></div>
                            <div class="flex space-x-2 py-1"><span class="w-4 h-4" style="background-color: rgb(101, 203, 243);">&nbsp;</span><span class="legendlabel">Confident
                                    (90 &gt; pLDDT &gt; 70)</span></div>
                            <div class="flex space-x-2 py-1"><span class="w-4 h-4" style="background-color: rgb(255, 219, 19);">&nbsp;</span><span class="legendlabel">Low (70 &gt;
                                    pLDDT &gt; 50)</span></div>
                            <div class="flex space-x-2 py-1"><span class="w-4 h-4" style="background-color: rgb(255, 125, 69);">&nbsp;</span><span class="legendlabel">Very low
                                    (pLDDT &lt; 50)</span></div>
                            <div class="row column legendDesc"> ESMFold produces a per-residue confidence
                                score (pLDDT) between 0 and 100. Some regions below 50 pLDDT may be unstructured in isolation.
                            </div>
                        </div>
                        <div class="w-1/2">
                            <div class="font-medium mt-4 flex items-center space-x-2"><b>Input structure </b><span class="w-4 h-4 bg-gray-300 inline-flex" ></span></div>
                            
                            <div class="flex space-x-2 py-1"><span class="w-4 h-4" style="background-color:hotpink" >&nbsp;</span><span class="legendlabel">Fixed positions</span></div>

                        </div>
                    </div>
            <script>

              function drawStructures(i, selectedResidues) {
            $("#rmsd").text(seqs[i]["rmsd"])
            $("#seqText").text(seqs[i]["seq"])
            $("#seqrmsd").text(seqs[i]["rmsd"])
            $("#id").text(i)
            $("#score").text(seqs[i]["score"])
            $("#recovery").text(seqs[i]["recovery"])
            $("#plddt").text(seqs[i]["plddt"])

            viewer = $3Dmol.createViewer(element, config);
            viewer.addModel(data[0], "pdb");
            viewer.addModel(pdb, "pdb");



            viewer.getModel(1).setStyle({}, { cartoon: { colorscheme: { prop: "resi", map: colors } } })
            viewer.getModel(0).setStyle({}, { cartoon: { colorfunc: colorAlpha } });
            viewer.zoomTo();
            viewer.render();
            viewer.zoom(0.8, 2000);
            viewer.getModel(0).setHoverable({}, true,
                function (atom, viewer, event, container) {
                    if (!atom.label) {
                        atom.label = viewer.addLabel(atom.resn + atom.resi + " pLDDT=" + atom.b, { position: atom, backgroundColor: "mintcream", fontColor: "black" });
                    }
                },
                function (atom, viewer) {
                    if (atom.label) {
                        viewer.removeLabel(atom.label);
                        delete atom.label;
                    }
                }
            );
        }
        let viewer = null;
        let voldata = null;
        let element = null;
        let config = null;
        let currentIndex = """
        + str(sequences[i]["Seq"])
        + """;
            let seqs = """
        + seqdata
        + """
               let data = """
        + pred_mol
        + """  
                let pdb = `"""
        + mol
        + """`  
         var selectedResidues = """
        + f"{selectedResidues}"
        + """
                //AlphaFold code from https://gist.github.com/piroyon/30d1c1099ad488a7952c3b21a5bebc96
                let colorAlpha = function (atom) {
                    if (atom.b < 0.50) {
                        return "OrangeRed";
                    } else if (atom.b < 0.70) {
                        return "Gold";
                    } else if (atom.b < 0.90) {
                        return "MediumTurquoise";
                    } else {
                        return "Blue";
                    }
                };
               
                let colors = {}
                for (let i=0; i<"""
        + str(num_res)
        + """;i++){
                if (selectedResidues.includes(i)){
                    colors[i]="hotpink"
                }else{
                    colors[i]="lightgray"
                }}

                let colorFixedSidechain = function(atom){
                                if (selectedResidues.includes(atom.resi)){
                                    return "hotpink"
                                }else if (atom.elem == "O"){
                                    return "red"
                                }else if (atom.elem == "N"){
                                    return "blue"
                                }else if (atom.elem == "S"){
                                    return "yellow"
                                }else{
                                    return "lightgray"
                                }
                            }

             $(document).ready(function () {
                element = $("#container");
                config = { backgroundColor: "white" };
                //viewer.ui.initiateUI();
 
                drawStructures(currentIndex, selectedResidues)
                $("#sidechain").change(function () {
                    if (this.checked) {
                        BB = ["C", "O", "N"]
                        
                        if ($("#startstructure").prop("checked")) {
                            viewer.getModel(0).setStyle( {"and": [{resn: ["GLY", "PRO"], invert: true},{atom: BB, invert: true},]},{stick: {colorscheme: "WhiteCarbon", radius: 0.3}, cartoon: { colorfunc: colorAlpha }});
                            viewer.getModel(1).setStyle( {"and": [{resn: ["GLY", "PRO"], invert: true},{atom: BB, invert: true},]},{stick: {colorfunc:colorFixedSidechain, radius: 0.3}, cartoon: {colorscheme:{prop:"resi",map:colors} }});
                        }else{
                            viewer.getModel(0).setStyle( {"and": [{resn: ["GLY", "PRO"], invert: true},{atom: BB, invert: true},]},{stick: {colorscheme: "WhiteCarbon", radius: 0.3}, cartoon: { colorfunc: colorAlpha }});
                            viewer.getModel(1).setStyle();                        
                        }
                        
                        viewer.render()
                    } else {
                        if ($("#startstructure").prop("checked")) {
                        viewer.getModel(0).setStyle({cartoon: { colorfunc: colorAlpha }});
                        viewer.getModel(1).setStyle({cartoon: {colorscheme:{prop:"resi",map:colors} }});
                        }else{
                            viewer.getModel(0).setStyle({cartoon: { colorfunc: colorAlpha }});
                            viewer.getModel(1).setStyle();
                            }
                        viewer.render()
                    }
                });
                $("#seq").change(function () {
                    drawStructures(this.value, selectedResidues)
                    currentIndex = this.value
                    $("#sidechain").prop( "checked", false );
                    $("#startstructure").prop( "checked", true );
                });
                $("#startstructure").change(function () {
                    if (this.checked) {
                         $("#sidechain").prop( "checked", false );
                       viewer.getModel(1).setStyle({},{cartoon: {colorscheme:{prop:"resi",map:colors} } })
                       viewer.getModel(0).setStyle({}, { cartoon: { colorfunc: colorAlpha } });
                       viewer.render()
                    } else {
                        $("#sidechain").prop( "checked", false );
                       viewer.getModel(1).setStyle({},{})
                       viewer.getModel(0).setStyle({}, { cartoon: { colorfunc: colorAlpha } });
                        viewer.render()
                    }
                });
                $("#download").click(function () {
                    download("outputs/esm_fold_prediction_"+currentIndex+".pdb", data[0]);
                })
        });
        function download(filename, text) {
            var element = document.createElement("a");
            element.setAttribute("href", "data:text/plain;charset=utf-8," + encodeURIComponent(text));
            element.setAttribute("download", filename);
            element.style.display = "none";
            document.body.appendChild(element);
            element.click();
            document.body.removeChild(element);
        }
        </script>
        </body></html>"""
    )

    return f"""<iframe style="width: 100%; height: 1300px" name="result" allow="midi; geolocation; microphone; camera; 
    display-capture; encrypted-media;" sandbox="allow-modals allow-forms 
    allow-scripts allow-same-origin allow-popups 
    allow-top-navigation-by-user-activation allow-downloads" allowfullscreen="" 
    allowpaymentrequest="" frameborder="0" srcdoc='{x}'></iframe>"""


def set_examples(example):
    (
        label,
        inp,
        designed_chain,
        fixed_chain,
        homomer,
        num_seqs,
        sampling_temp,
        atomsel,
    ) = example
    return [
        label,
        inp,
        designed_chain,
        fixed_chain,
        homomer,
        gr.Slider.update(value=num_seqs),
        gr.Radio.update(value=sampling_temp),
        atomsel,
    ]


import hashlib


def get_esmfold(sequence):
    headers = {
        "Content-Type": "application/x-www-form-urlencoded",
    }
    sequence = sequence.replace("/", ":")
    filename = "outputs/" + hashlib.md5(str.encode(sequence)).hexdigest() + ".pdb"

    if not os.path.exists(filename):
        response = requests.post(
            "https://api.esmatlas.com/foldSequence/v1/pdb/",
            headers=headers,
            data=sequence,
            verify=False
        )

        name = sequence[:3] + sequence[-3:]
        pdb_string = response.content.decode("utf-8")

        with open(filename, "w+") as f:
            f.write(pdb_string)
        print("retrieved prediction", filename)
    else:
        print("prediction already on disk")
    return filename


proteinMPNN = gr.Blocks()

with proteinMPNN:
    gr.Markdown("# ProteinMPNN + ESMFold")
    gr.Markdown(
        """This model takes as input a protein structure and based on its backbone predicts new sequences that will fold into that backbone. 
        It will then run [ESMFold](https://esmatlas.com/about) by MetaAI on the predicted structures and align the predicted structure for the designed sequence with the original backbone. 

        **Note, there is a 400 residue limit in this version and multimeric structures can only be predicted locally. Follow, [README](https://huggingface.co/spaces/simonduerr/ProteinMPNNESM/blob/main/README.md) for instructions on how to run locally.** 
        """
    )

    with gr.Tabs():
        with gr.TabItem("Input"):
            inp = gr.Textbox(
                placeholder="PDB Code or upload file below", label="Input structure"
            )
            file = gr.File(file_count="single")

        with gr.TabItem("Settings"):
            with gr.Row():
                designed_chain = gr.Textbox(value="A", label="Designed chain")
                fixed_chain = gr.Textbox(
                    placeholder="Use commas to fix multiple chains", label="Fixed chain"
                )
            with gr.Row():
                num_seqs = gr.Slider(
                    minimum=1, maximum=15, value=1, step=1, label="Number of sequences" 
                )
                sampling_temp = gr.Radio(
                    choices=[0.1, 0.15, 0.2, 0.25, 0.3],
                    value=0.1,
                    label="Sampling temperature",
                    
                )
            gr.Markdown(
                """ Sampling temperature for amino acids, `T=0.0` means taking argmax, `T>>1.0` means sample randomly. Suggested values `0.1, 0.15, 0.2, 0.25, 0.3`. Higher values will lead to more diversity.
            """
            )
            with gr.Row():
                model_name = gr.Dropdown(
                    choices=[
                        "v_48_002",
                        "v_48_010",
                        "v_48_020",
                        "v_48_030",
                    ],
                    label="Model",
                    value="v_48_020",
                    
                )
                backbone_noise = gr.Dropdown(
                    choices=[0, 0.02, 0.10, 0.20, 0.30], label="Backbone noise", value=0, 
                    
                )
            with gr.Row():
                homomer = gr.Checkbox(value=False, label="Homomer?")
                gr.Markdown(
                    "for correct symmetric tying lenghts of homomer chains should be the same"
                )
            gr.Markdown("## Fixed positions")
            gr.Markdown(
                """You can fix important positions in the protein. Resid should be specified with the same numbering as in the input pdb file. The fixed residues will be highlighted in the output. 
                The [VMD selection](http://www.ks.uiuc.edu/Research/vmd/vmd-1.9.2/ug/node89.html) synthax is used. You can also select based on ligands or chains in the input structure to specify interfaces to be fixed.

                 - <code>within 5 of resid 94</code> All residues that have >1 atom closer than 5 Å to any atom of residue 94
                 - <code>name CA and within 5 of resid 94</code> All residues that have CA atom closer than 5 Å to any atom of residue 94
                 - <code>resid 94 96 119</code> Residues 94, 94 and 119
                 - <code>within 5 of resname ZN</code> All residues with any atom <5 Å of zinc ion
                 - <code>chain A and within 5 of chain B </code> All residues of chain A that are part of the interface with chain B
                 - <code>protein and within 5 of nucleic </code> All residues that bind to DNA (if present in structure)
                 - <code>not (chain A and within 5 of chain B) </code>  only modify residues that are in the interface with the fixed chain, not further away
                 - <code>chain A or (chain B and sasa < 20) </code> Keep chain A and all core residues fixeds
                 - <code>pLDDT >70 </code> Redesign all residues with low pLDDT
                 
                 Note that <code>sasa</code> and <code>pLDDT</code> selectors modify default VMD behavior. SASA is calculated using moleculekit and written to the mass attribute. Selections based on mass do not work. 
                 pLDDT is an alias for beta, it only works correctly with structures that contain the appropriate values in the beta column of the PDB file. """
            )
            atomsel = gr.Textbox(
                placeholder="Specify atom selection ", label="Fixed positions", 
                api_name= "fixed_positions"
            )

        btn = gr.Button("Run")
    label = gr.Textbox(label="Label", visible=False)

    samples = [["Monomer design", "6MRR", "A", "", False, 2, 0.1, ""]]

    if is_local:
        samples.extend(
            [
                ["Homomer design", "1O91", "A,B,C", "", True, 2, 0.1, ""],
                [
                    "Redesign of Homomer to Heteromer",
                    "3HTN",
                    "A,B",
                    "C",
                    False,
                    2,
                    0.1,
                    "",
                ],
                [
                    "Redesign of MID1 scaffold keeping binding site fixed",
                    "3V1C",
                    "A,B",
                    "",
                    False,
                    2,
                    0.1,
                    "within 5 of resname ZN",
                ],
                [
                    "Redesign of DNA binding protein",
                    "3JRD",
                    "A,B",
                    "",
                    False,
                    2,
                    0.1,
                    "within 8 of nucleic",
                ],
                [
                    "Surface Redesign of miniprotein",
                    "7JZM",
                    "A,B",
                    "",
                    False,
                    2,
                    0.1,
                    "chain B or (chain A and sasa < 20)",
                ],
            ]
        )
    examples = gr.Dataset(
        components=[
            label,
            inp,
            designed_chain,
            fixed_chain,
            homomer,
            num_seqs,
            sampling_temp,
            atomsel,
        ],
        samples=samples,
    )

    gr.Markdown("# Output")

    with gr.Tabs():
        with gr.TabItem("Designed sequences"):
            chosen_seq = gr.Dropdown(
                choices=[],
                label="Select a sequence for validation",
            )
            mol = gr.HTML()
            out = gr.Textbox(label="Fasta Output")

        with gr.TabItem("Amino acid probabilities"):
            plot = gr.Plot()
            all_log_probs = gr.File(visible=False)
        with gr.TabItem("T adjusted probabilities"):
            gr.Markdown("Sampling temperature adjusted amino acid probabilties")
            plot_tadjusted = gr.Plot()
            all_probs = gr.File(visible=False)

    tempFile = gr.Variable()
    selectedResidues = gr.Variable()
    seq_dict = gr.Variable()
    btn.click(
        fn=update,
        inputs=[
            inp,
            file,
            designed_chain,
            fixed_chain,
            homomer,
            num_seqs,
            sampling_temp,
            model_name,
            backbone_noise,
            atomsel,
        ],
        outputs=[
            out,
            plot,
            plot_tadjusted,
            all_log_probs,
            all_probs,
            tempFile,
            chosen_seq,
            selectedResidues,
            seq_dict,
            mol,
        ],
        api_name = "proteinmpnn"
    )
    chosen_seq.change(
        updateseq,
        inputs=[chosen_seq, seq_dict, tempFile, selectedResidues],
        outputs=mol,
    )
    examples.click(fn=set_examples, inputs=examples, outputs=examples.components)
    gr.Markdown(
        """Citation: **Robust deep learning based protein sequence design using ProteinMPNN** <br>
Justas Dauparas, Ivan Anishchenko, Nathaniel Bennett, Hua Bai, Robert J. Ragotte, Lukas F. Milles, Basile I. M. Wicky, Alexis Courbet, Robbert J. de Haas, Neville Bethel, Philip J. Y. Leung, Timothy F. Huddy, Sam Pellock, Doug Tischer, Frederick Chan, Brian Koepnick, Hannah Nguyen, Alex Kang, Banumathi Sankaran, Asim Bera, Neil P. King, David Baker <br>
Science Vol 378, Issue 6615, pp. 49 -56; doi: [10.1126/science.add2187](https://doi.org/10.1126/science.add2187 <br><br> Server built by [@simonduerr](https://twitter.com/simonduerr) and hosted by Huggingface"""
    )


proteinMPNN.launch()