EN KR
Linkedin Facebook

SaaS

Auto-BP (Binding Pose)

We introduce an advanced approach (auto-Best pose) to predict the most likely binding poses of ligands within target proteins, which analyzes binding modes while accounting the integration of structural information, consideration of ligand and protein flexibility, and reliable scoring and evaluation metrics. A fully automated CLOUD SaaS process that can improve efficiency, consistency, scalability, accessibility, and integration with AI algorithms are significant contribution to modern drug discovery efforts. It plays a valuable role for researchers and pharmaceutical companies in identifying promising drug candidates effectively and swiftly.

We introduce an advanced approach (auto-Best pose) to predict the most likely binding poses of ligands within target proteins, which analyzes binding modes while accounting the integration of structural information, consideration of ligand and protein flexibility, and reliable scoring and evaluation metrics. A fully automated CLOUD SaaS process that can improve efficiency, consistency, scalability, accessibility, and integration with AI algorithms are significant contribution to modern drug discovery efforts. It plays a valuable role for researchers and pharmaceutical companies in identifying promising drug candidates effectively and swiftly.
SCROLL DOWN

Solution

Researchers found a hit and its binding protein through experiments including cell-based essays, and many other methods, but they cannot proceed its lead optimization because they do not know how it binds.

Auto-BP service is appropriate. For Auto-BP, it is recommended to select low, intermediate or advanced levels depending on the importance of the customer’s lead compound.

A molecular researchers found it certain that it plays a role in cell death, but it is unknown which protein it binds to. What service is appropriate in this case?

MOA (mode of action) service is appropriate. In this case, Auto-BP is set lightly, and instead of performing Auto-BP one molecule against one target 2,000 panel targets are screened simultaneously and a list of suitable candidates is givenas an output. (https://cloud.syntekabio.com)

Theory behind Binding Pose Prediction
Software

The AI algorithms of our precision in-house scoring system, are key components underline our screening protocol enabling accurate prediction of binding affinity and orientation. The platform has fully automated all processes from the maximum collection of molecular docking to 3D-CNN and MD simulation for enhanced accuracy and performance in scoring and ranking.

Pocket model (~100 templates)

APocket model (~100 templates)

3D-CNN Model

B3D-CNN Model

MD-simulation, Strain & Binding Free Energy

CMD-simulation, Strain & Binding
Free Energy

Auto-BP Accuracy

DAuto-BP Accuracy

  • Theoretically, if an infinite number of poses are created, and 3D-CNN consistently selects the Best-pose among them, and if their plots are thermodynamically meaningful like a sigmoid shape, it can be said that the top pose is close to the best pose.
  • (A) If about 100 template molds are created in a protein pocket and 1,000 poses are created including conformers for each docking template, 100,000 poses are generated as shown to be a sigmoid shape in (B) If we extract the top 100 and draw superimpose images, it looks like the following (C) And, this is a var graph that evaluates how many crystal structure-level pose appear when top 2 are selected blindly using about 200 crystal structures whose poses are known. When simple docking, Auto-BP, and Auto-BP-MD are used, it gives almost 94% performance. (D) The dark blue is the competitor’s data, but it is not disclosed here.

Application

  • Best-pose screening is useful for structure-based drug discovery. This can be applied at various stages of the drug discovery and development process for HIT screening(1), enhancing efficacy from HIT to Lead generation and optimization by QSAR analysis utilizing binding mode prediction(2), and Off-Target discovery and indication expansion such as all for 2000 panels, G for GPCR panels, K for Kinase panels(3). STB software suites for SaaS can be utilized to (1) Auto-HIT, (2) ALGO, and (3) OFF-All, G and K in STB CLOUD
References:
  • Ragoza, Matthew, etal. "Protein–ligand scoring with convolutional neural networks."Journal of chemical information and modeling 57.4 (2017): 942-957
  • Syntekabio’s Hit to Preclinical Candidate on CLOUD Computing, Bio PharmaTrend, Vol4. 2023
  • How Syntekabio’s Supercomputing Enables AI-driven Drug Discovery, Bio PharmaTrend, Vol3. 2023

About Auto-Bestpose(BP) Solution

About Auto-Bestpose(BP) Service

Auto-Bestpose(BP) is a software tool to predict the interaction between key residues present in the binding site of the target protein and the ligand. Through pose-analysis of the binding mode of disease-causing proteins and ligands, it provides optimized best pose to improve the potency of the drug compounds and optimize hits to leads.
Auto-BP program can be applied for discovering HIT discovery, Lead optimization, Off-Target discovery and indication expansion.
Auto-Bestpose image

Auto-Bestpose

  • Download structural domain in PDB format
  • See structural domain only (PV viewer)

Application

  • HIT screening, HIT to Lead, Lead optimization, Off-Target discovery and indication expansion

Key models and methods including

  • Mold generation (100 multi-template)
  • Docking pocket identification (factors in pocket environment)
  • Scoring with 3D-CNN machine learning
  • Molecular conformation generation (1K conformers)
  • Calculations of binding free energy (B.E.)
  • Protein backbone RMSD analysis (RMSD)
  • Key residue analysis (3D structure-based and molecular dynamics simulation)
  • Molecular dynamics simulation

Deliverables

  • Two candidates predicted as Bestpose Characteristic information including physicochemical properties of target-ligand complex

Advantages of our auto-BPservices:

Our high-performance bio-supercomputing processes, synergistically integrated with our cutting-edge SaaS software, proficiently manage extensive calculations with heightened precision. Our software facilitates the execution of high-quality customized projects, encompassing advanced computing simulations and professional drug design, spanning from preliminary assessments to result validation.

  • Service
  • Tool
  • Module
  • Datasource

  • Bestpose SCR

    Auto-BP

    Docking-module

    STB Virtual library

  • Zinc SCR (1B)

    DMC-Zinc-Hit

    SCR-Docking-module

    1B Zinc library

  • LFS HIT-SCR (50M)

    DMC-QSL-Hit

    SCR-Docking-module

    50M Synplelibrary

  • FDA drug SCR (10K)

    DMC-DR-Hit

    SCR-Docking-module

    10K FDA drug

  • Hit-MOA-Kinases

    600-Kinase-profile

    SCR-Docking-module

    Kinase targets

  • Hit-MOA-Membrane

    400-Mem-profile

    SCR-Docking-module

    Membrane targets

  • Hit-MOA-Targets

    2000-Target-profile

    SCR-Docking-module

    All known targets

About Auto-Bestpose(BP) Solution

Accurate prediction of target-to-ligand interactions plays a critical role in the identification of ideal therapeutic compounds by assisting structure-based drug design. The auto-BP service, which predicts binding modes and affinity, predicts the best pose of target and ligand with 3D docking, 3D-CNN machine learning binding tuning based on physicochemical mechanisms, and provides further predictive accuracy of binding modes through MD simulations.
In-house key-residue analysis algorithms can determine and predict which partners are strong or weak in interaction with an accurate calculation of specific binding sites and binding affinity. By docking and scoring of auto-BP, the ligand with the most stable binding energy and the best affinity to the receptor is selected. Our prediction accuracy has achieved a superior success rate compared to other programs; see our performance data.
Auto-BP Accuracy image

Auto-BP Accuracy

J. Chem. Theory Comput. 2021,17,4 2630-2639

  • Schrodinger
  • Syntekabio

Coretechnology

  • Better sampling with multiple templates using Mold-generation
  • ENVA docking prediction based on scoring function of physicochemical properties.
  • Fine tunning with 3D-CNN (convolutional neural network) machine learning
  • Deep learning-based best-pose prediction using more than 40,000 conformers
  • In-silico evaluation through automated molecular dynamics (MD) simulation
Binding pose
  • Binding
    pose
  • RMSD
    distribution
  • Gapdock PMSD
    <2.5
  • Conformer PMSD
    <2.5
  • Single
    template
    0%
    (0/12 poses)
    ~0%
    (2/60000 poses)
  • Multiple
    templates
    17.8%
    (89/500 poses)
    7.95%
    (3162/39484 poses)
  • RMSD < 2.5
  • RMSD =< 2.5

Best-pose

This program is in-silico prediction result, so werecommend proceeding with next-stage experiments such as in-vitro validationtests with these Best-pose candidates

Binding pose

Identify the best-pose to predict the Interaction between target and ligand. Main processes run automatically from screening to 3D-CNN & MD simulation. It generates more than 40,000 conformers as potential poses, select the best-pose

Process

  • HIT Screening
  • Physical docking
  • 3D-CNN machine learning
  • Select the top 1,000 compounds
  • Generate more than 40,000 conformers of target andligand
  • Select the top 1,000 best-pose candidates
  • Run the MD simulation with the top 1,000 
  • Identify the best-pose candidate

Input

Target +ligand

Runtime

2~3 weeks per 1 run

Output

10 Best-pose with
B.E. value

Support

Computer capacity :
I Unit of 100 CPU
/100 GPU

send us your inquiry

SUBMIT
SUBMIT