Case Study
Off-Target Risk Prediction Service

Inquiry
Off-Target Risk Prediction Service - CD ComputaBio
Selectivity & Safety Target Profiling

Off‑Target Risk
Prediction Service

Map plausible unintended targets, connect them with safety‑relevant biology, and prioritize the counter‑screens that can most efficiently de‑risk a lead series.

Broaden
Prioritize
Verify
Decide
Our Services

Off‑Target Risk Prediction Services

Select a focused module or combine them as an investigation that moves from an unexplained signal to a compact, justified counter-screen.

Hypothesis‑first project design

The workflow begins with the safety question or selectivity decision, not with a predefined modeling package.

01
Reveal Hidden Pharmacology

Ligand‑ and Chemogenomic Off‑Target Search

Find plausible unintended targets using known ligand sets, bioactivity relationships, and multi‑task prediction. Target hypotheses, chemical evidence, coverage flags.

02
Add Structural Context

Focused Structure‑Based Screening

Use reverse docking or binding‑site modeling for suitable targets to provide orthogonal evidence, not a universal proteome score. Pose hypotheses, site compatibility, structure QC.

03
Connect to Safety

Target‑to‑Pathway and Phenotype Mapping

Prioritize interactions that are biologically and exposure relevant rather than ranking targets by model score alone. Safety‑target annotation, tissue/pathway context, ADR hypotheses.

04
Test the Right Targets

Counter‑Screen Panel Design

Translate computational hypotheses into a concise validation panel that can support advance, redesign, or stop decisions. Assay priorities, comparator controls, decision thresholds.

05
Selectivity Optimization

Analog Series Promiscuity Comparison

Evaluate off‑target profiles across a chemical series to identify analogs with improved selectivity without sacrificing primary activity.

06
Unexpected Phenotype

Adverse Effect Mechanism Hypotheses

Connect phenotypic observations with plausible target interactions and propose discriminatory experiments to confirm or reject the mechanism.

07
Data Integration

Custom Target Panel Modeling

Incorporate proprietary counter‑screen data into a project‑specific scoring model with assay‑aware curation and validation.

Hypothesis Funnel

Start wide. Finish with a testable shortlist.

An off-target search may nominate hundreds of proteins. The useful work is the controlled narrowing: removing weak associations, adding tissue and exposure context, and preserving only hypotheses that can change a development decision.

Each funnel stage records why a target was retained or removed, producing an auditable counter-screen rationale rather than an unexplained ranked list.

Counter-screen shortlist
Cast the hypothesis net

Combine ligand, chemogenomic, binding-site, and known pharmacology evidence.

Remove implausible biology

Filter by tissue access, expression, pathway relevance, and achievable free exposure.

Challenge the evidence

Compare orthogonal support, uncertainty, assay artifacts, and analog-series consistency.

Build the smallest useful panel

Select discriminatory targets, controls, concentration ranges, and stop/go criteria.

Challenge‑Driven Design

Investigate the mechanism behind an unwanted signal

Begin with the phenotype, organ concern, or selectivity gap—not with a generic target panel—and work backward to the evidence needed.

Signals that trigger an investigation

01
Which unintended targets should we test first?

Orthogonal evidence and exposure‑weighted ranking to build a focused counter‑screen panel.

02
Why did an unexpected phenotype appear?

Target‑pathway‑phenotype hypothesis mapping with discriminatory experiments.

03
Which analog is most selective?

Series‑level off‑target and promiscuity comparison with selectivity tiers.

04
Can reverse docking add value?

Target‑structure suitability and focused reverse screening with structure‑supported hypotheses.

05
How to prioritize redesign vs. confirm?

Evidence strength + biological context to select the most informative next step.

Evidence assembled for each signal

Each response is chosen to discriminate among plausible mechanisms, reduce an oversized target list, and specify the next counter-screen.

Hypothesis Generation Ligand‑set similarity, chemogenomic models, and multi‑task machine learning.
Structural Evidence Reverse docking, binding‑site compatibility, and structure‑based panels.
Biological Context Pathway, tissue, safety‑target, and adverse‑event annotations.
Exposure Plausibility Free concentration, target affinity, and distribution filters.
Panel Design Counter‑screen prioritization, controls, and decision thresholds.
Target Dossiers

Organize candidates by why they matter

A protein earns a place on the shortlist for a specific reason. These evidence dossiers keep hazard, biology, exposure, and structural similarity from collapsing into one opaque score.

01

Documented hazard

Proteins linked with cardiac, CNS, hepatic, or other organ‑system adverse effects.

Dossier basis: adverse-outcome evidence
02

Promiscuous families

Groups known for promiscuity or off‑target liabilities (e.g., GPCRs, kinases, ion channels).

Dossier basis: family-wide binding history
03

Mechanistic neighbors

Targets within the same biological pathways as the intended target.

Dossier basis: pathway proximity
04

Exposure-accessible tissue

Targets with high expression in relevant organs (heart, liver, brain, etc.).

Dossier basis: expression × distribution
05

Binding-site mimics

Proteins with similar binding sites to the intended target.

Dossier basis: pocket resemblance
Computational Platform

Connect chemistry and biology to safety decisions

Off‑target prediction spans ligand‑based similarity, chemogenomic models, structure‑based reverse screening, and biological context integration. Our platform combines these to provide ranked, evidence‑qualified hypotheses.

Each prediction is accompanied by the type and strength of supporting evidence, applicability flags, and a clear path to experimental validation.

Staged prediction reduces false confidence

Broad hypotheses are filtered by biological relevance and experimental feasibility before being promoted to the counter‑screen panel.

01
Ligand‑Set Similarity

Compare compound structures with known target ligand sets to identify unexpected associations.

02
Chemogenomic Models

Multi‑task machine learning trained on bioactivity data across target families.

03
Reverse Docking

Fit compounds into target binding sites for suitable structures to provide orthogonal evidence.

04
Biological Annotation

Pathway, tissue, safety‑target, and adverse‑event mapping for each candidate target.

05
Evidence Integration

Combine scores, confidence, exposure context, and program criteria into a ranked panel.

Representative Project Scenarios

Choose the investigation pattern that matches the signal

Each scenario begins with a different clue—selectivity loss, unexpected phenotype, analog drift, or proprietary evidence—and therefore demands a different path through target space.

Safety Target Triage

Prioritize counter‑screens before lead nomination.

Target universe → orthogonal prediction → exposure‑weighted panel with evidence types and confidence flags.

Output: focused counter‑screen panel and validation rationale.
Unexpected Phenotype Review

Connect a cellular signal with plausible unintended targets.

Phenotype context → target mapping → pathway analysis → discriminatory assay proposals.

Output: mechanism hypotheses and experimental plan.
Selectivity Optimization

Compare analog promiscuity and redesign options.

Series scoring → evidence ladder → chemistry priorities with structural modifications.

Output: selectivity tiers and redesign recommendations.
Custom Panel Modeling

Incorporate proprietary counter‑screen data.

Assay‑aware curation → model validation → project‑specific scoring and calibration.

Output: validated internal model and decision framework.
Project Workflow

From safety question to decision‑ready report

The investigation is configured around the observed signal, accessible target evidence, exposure assumptions, and the counter-screen threshold that will change the program's direction.

01

Define the Question

Identify the safety decision: triage, phenotype, selectivity, or panel design.

02

Curate Data

Structures, intended target, known pharmacology, exposure context, and program criteria.

03

Run Predictions

Apply ligand‑based, chemogenomic, and structure‑based methods to generate hypotheses.

04

Integrate Evidence

Combine prediction strength, biological relevance, exposure plausibility, and confidence.

05

Deliver Decision

Provide ranked panel, experimental plan, and redesign or confirmation priorities.

Project Deliverables

Results designed for development decisions

The final package combines ranked target hypotheses, biological context, confidence annotations, and a focused validation plan.

01 Ranked Target Matrix

Compound‑target associations with evidence types and scores.

02 Confidence Annotations

Evidence strength, applicability flags, data provenance, and uncertainty notes.

03 Biological Context

Pathway, tissue, safety‑target, and potential adverse‑effect mapping.

04 Selectivity Comparison

Analog‑series promiscuity and selectivity tiers.

05 Counter‑Screen Panel

Prioritized targets, controls, assay recommendations, and decision criteria.

06 Technical Package

Models, datasets, structures, plots, and documentation.

Frequently Asked Questions

Planning an off‑target risk project

Does a predicted off‑target prove that a compound will cause an adverse effect?

No. Binding or activity must be confirmed, and clinical relevance depends on potency, free exposure, tissue distribution, pathway context, and compensatory biology.

How large should the target panel be?

The useful panel is decision‑dependent. Broad virtual screening can generate hypotheses, while a focused experimental panel should prioritize high‑evidence, exposure‑plausible, safety‑relevant targets.

Can reverse docking screen the whole proteome?

Proteome‑scale docking is limited by structure availability, binding‑site definition, scoring bias, protein flexibility, and compute. It is best used selectively and combined with ligand and bioactivity evidence.

Can peptides or biologics be assessed?

Yes, but the evidence sources, interaction models, and target universe differ from small molecules and should be scoped specifically for the modality.

How are unknown chemotypes handled?

Predictions with weak ligand support, poor model applicability, or unsuitable target structures are flagged as uncertain and are not promoted solely by a high score.

What information is needed to start a project?

Helpful inputs include structures, intended target, known pharmacology, exposure context, safety concerns, and the decision the study should support.

Which off‑targets matter for your program?

Send the structures, intended target, any unexpected assay or phenotype results, and the exposure range. CD ComputaBio will define a focused off-target investigation and validation panel.

Start Your Project

Online Inquiry

Submit your project details below, and our team will respond within 24 hours.

x
Need help getting the data you need?

Talk to our technical team about your project!

I Want To Talk