SkillAgentSearch skills...

medchem

Medicinal chemistry filters for compound triage. Drug-likeness rules (Lipinski Ro5, Veber, Oprea, CNS, leadlike, REOS, Golden Triangle, Ro3), structural alerts (PAINS, NIBR, Lilly Demerits), chemical group detectors, complexity metrics, and filter composition query language. Built on RDKit/datamol.

Install / Use

npx skills add jaechang-hits/SciAgent-Skills --skill medchem

Installs into whichever agent you are using.

About this skill
📄

SKILL.md

Installable skill definition

Quality Score

91/100

Supported Platforms

Universal

Our assessment of medchem

medchem scores 91/100 on our quality scale, 1169th of 4,716 Development & Engineering skills we index (top 25%).

Its SKILL.md is 17 KB long, well organised into 63 sections with 17 code examples: a thorough specification that gives an agent plenty to work with.

It has 367 GitHub stars, a meaningful sign that others use it.

Substance
30/30
Structure
20/20
Description
15/15
Adoption
11/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated about 3 months ago, so medchem is actively maintained.
  • Our last check on 2026-11-20 found the source still online.
  • No license is declared. By default that means all rights are reserved: you can read it, but reusing or redistributing it is not clearly permitted. Ask the author before building on it commercially.
  • Its trust signals score 88/100, with 1 caution from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.

AI review by kimi-k2.7-code on 2026-10-07. Automated pattern scan on 2026-10-07. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

medchem compared with similar skills

All 4 of these similar skills score higher than medchem; compare them before choosing.

SkillScoreStarsUpdatedFormat
medchem (this skill)by jaechang-hits913673mo agoSKILL.md
taste-skillby Leonxlnx100118.9k11d agoCLAUDE.md
Agent-Reachby Panniantong100113.3k12d agoCLAUDE.md
hyperframesby heygen-com10087.1ktodayCLAUDE.md
headroomby headroomlabs-ai10080.2ktodayCLAUDE.md

Frequently asked questions

How do I install medchem?
Run npx skills add jaechang-hits/SciAgent-Skills --skill medchem. The install tabs above show the steps for each supported agent.
Which AI agents does medchem work with?
It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
Is medchem safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It declares no license and scores 88/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is medchem still maintained?
The repository was last updated about 3 months ago, so medchem is actively maintained.

name: medchem description: >- Medicinal chemistry filters for compound triage. Drug-likeness rules (Lipinski Ro5, Veber, Oprea, CNS, leadlike, REOS, Golden Triangle, Ro3), structural alerts (PAINS, NIBR, Lilly Demerits), chemical group detectors, complexity metrics, and filter composition query language. Built on RDKit/datamol. For hit-to-lead filtering, library design, ADMET pre-screening. For molecular I/O use rdkit-cheminformatics or datamol. license: Apache-2.0

Medchem

Overview

Medchem is a Python library for molecular filtering and prioritization in drug discovery. It provides hundreds of established medicinal chemistry rules, structural alerts, and chemical group detectors to triage compound libraries at scale. All filters support parallel execution and return structured results.

When to Use

  • Applying drug-likeness rules (Lipinski, Veber, Oprea, CNS, REOS) to compound libraries
  • Filtering molecules by structural alerts (PAINS, NIBR, Lilly Demerits)
  • Detecting specific chemical groups (hinge binders, Michael acceptors, reactive groups)
  • Calculating molecular complexity metrics (Bertz, Whitlock, Barone)
  • Applying custom property constraints (MW, LogP, TPSA, rotatable bonds)
  • Composing complex multi-rule filter queries with Boolean logic
  • For SMILES/SDF parsing, descriptors, and fingerprints use rdkit-cheminformatics
  • For high-level molecular manipulation use datamol-cheminformatics

Prerequisites

pip install medchem datamol

Medchem depends on RDKit and datamol. All molecule inputs are RDKit Chem.Mol objects; use datamol.to_mol() to convert from SMILES.

Quick Start

import datamol as dm
import medchem as mc

# Convert SMILES to molecules
smiles_list = ["CC(=O)OC1=CC=CC=C1C(=O)O", "c1ccccc1N", "O=C(O)c1ccccc1"]
mols = [dm.to_mol(s) for s in smiles_list]

# Apply Rule of Five + structural alerts in one pass
rule_filter = mc.rules.RuleFilters(rule_list=["rule_of_five"])
alert_filter = mc.structural.CommonAlertsFilters()

rule_results = rule_filter(mols=mols, n_jobs=-1)
alert_results = alert_filter(mols=mols, n_jobs=-1)

print(f"Rule results: {rule_results}")
print(f"Alert results: {[r['has_alerts'] for r in alert_results]}")

Core API

1. Drug-Likeness Rules

Apply established medicinal chemistry rules via mc.rules. Individual rules return bool; RuleFilters applies multiple rules in batch.

import medchem as mc

# Single rule on a SMILES string
passes = mc.rules.basic_rules.rule_of_five("CC(=O)OC1=CC=CC=C1C(=O)O")
print(f"Passes Ro5: {passes}")  # True

# Available individual rules:
# rule_of_five, rule_of_three, rule_of_oprea, rule_of_cns,
# rule_of_leadlike_soft, rule_of_leadlike_strict, rule_of_veber,
# rule_of_reos, rule_of_drug, golden_triangle, pains_filter
import datamol as dm
import medchem as mc

# Batch application with RuleFilters
mols = [dm.to_mol(s) for s in smiles_list]
rfilter = mc.rules.RuleFilters(
    rule_list=["rule_of_five", "rule_of_oprea", "rule_of_cns"]
)
results = rfilter(mols=mols, n_jobs=-1, progress=True)
# Returns list of dicts: [{"rule_of_five": True, "rule_of_oprea": False, ...}, ...]
print(f"First molecule: {results[0]}")

2. Structural Alert Filters

Detect problematic structural patterns via mc.structural. Three filter sets cover different scope and stringency.

import datamol as dm
import medchem as mc

mol = dm.to_mol("c1ccc(N)cc1")
mols = [dm.to_mol(s) for s in smiles_list]

# Common Alerts — general structural alerts from ChEMBL / literature
alert_filter = mc.structural.CommonAlertsFilters()
has_alerts, details = alert_filter.check_mol(mol)  # single molecule
batch_results = alert_filter(mols=mols, n_jobs=-1, progress=True)
# Each result: {"has_alerts": bool, "alert_details": [...], "num_alerts": int}
print(f"Alerts: {batch_results[0]}")
import medchem as mc

# NIBR Filters — Novartis industrial filter set (returns bool list)
nibr_filter = mc.structural.NIBRFilters()
nibr_results = nibr_filter(mols=mols, n_jobs=-1)
print(f"NIBR pass: {nibr_results}")  # [True, False, ...]

# Lilly Demerits — 275 patterns, molecules rejected at >100 demerits
lilly_filter = mc.structural.LillyDemeritsFilters()
lilly_results = lilly_filter(mols=mols, n_jobs=-1)
# Each result: {"demerits": int, "passes": bool, "matched_patterns": [...]}
print(f"Lilly: {lilly_results[0]}")

3. Chemical Groups

Detect specific functional group motifs via mc.groups.ChemicalGroup.

Predefined groups: hinge_binders, phosphate_binders, michael_acceptors, reactive_groups.

import medchem as mc

# Check for kinase hinge binders and Michael acceptors
group = mc.groups.ChemicalGroup(
    groups=["hinge_binders", "michael_acceptors"]
)

has_matches = group.has_match(mols)        # List[bool]
match_info = group.get_matches(mols[0])    # {group_name: [(atom_indices), ...]}
all_matches = group.get_all_matches(mols)  # List[Dict]
print(f"Has hinge binder: {has_matches}")

# Custom SMARTS patterns
custom = mc.groups.ChemicalGroup(
    groups=["reactive_groups"],
    custom_smarts={"trifluoromethyl_ketone": "[C;H0](=O)C(F)(F)F"}
)

4. Named Catalogs

Access curated chemical structure catalogs via mc.catalogs.

Available catalogs: functional_groups, protecting_groups, reagents, fragments.

import medchem as mc

catalog = mc.catalogs.NamedCatalogs.get("functional_groups")
matches = catalog.get_matches(mol)
print(f"Functional group matches: {matches}")

5. Molecular Complexity

Calculate synthetic accessibility proxies via mc.complexity.

Methods: bertz (topological), whitlock, barone.

import datamol as dm
import medchem as mc

mol = dm.to_mol("CC(=O)OC1=CC=CC=C1C(=O)O")

# Single molecule complexity
score = mc.complexity.calculate_complexity(mol, method="bertz")
print(f"Bertz complexity: {score:.1f}")

# Batch filtering by complexity threshold
cfilter = mc.complexity.ComplexityFilter(max_complexity=500, method="bertz")
results = cfilter(mols=mols, n_jobs=-1)
print(f"Passes complexity: {results}")  # List[bool]

6. Property Constraints

Apply custom property-based constraints via mc.constraints.Constraints.

import medchem as mc

constraints = mc.constraints.Constraints(
    mw_range=(200, 500),
    logp_range=(-2, 5),
    tpsa_max=140,
    rotatable_bonds_max=10,
    hbd_max=5,
    hba_max=10,
    rings_range=(1, 5),
    aromatic_rings_max=3,
)
results = constraints(mols=mols, n_jobs=-1)
# Each result: {"passes": bool, "violations": ["mw_range", ...]}
print(f"Violations: {results[0]}")

7. Query Language

Compose complex filter logic with Boolean expressions via mc.query.

import medchem as mc

# Parse a query combining rules, alerts, and property checks
query = mc.query.parse("rule_of_five AND NOT common_alerts")
results = query.apply(mols=mols, n_jobs=-1)  # List[bool]
print(f"Passing: {sum(results)}/{len(results)}")

# More complex queries
q2 = mc.query.parse("rule_of_cns AND complexity < 400")
q3 = mc.query.parse("(rule_of_five OR rule_of_oprea) AND NOT pains_filter")
q4 = mc.query.parse("mw > 200 AND mw < 500 AND logp < 5")

8. Functional API & Utilities

Shortcut functions in mc.functional and utilities in mc.utils.

import medchem as mc

# Functional API — one-liner filters
nibr_ok = mc.functional.nibr_filter(mols=mols, n_jobs=-1)    # List[bool]
alerts = mc.functional.common_alerts_filter(mols=mols, n_jobs=-1)
lilly = mc.functional.lilly_demerits_filter(mols=mols, n_jobs=-1)

# Utilities
standardized = mc.utils.standardize_mol(mol)  # sanitize, neutralize charges
batch_out = mc.utils.batch_process(
    mols=mols, func=mc.complexity.calculate_complexity,
    n_jobs=-1, progress=True, batch_size=100
)

Key Concepts

Rule Thresholds Quick Reference

| Rule | MW | LogP | HBD | HBA | RotBonds | TPSA | Other | |------|----|----- |-----|-----|----------|------|-------| | Ro5 (Lipinski) | ≤500 | ≤5 | ≤5 | ≤10 | — | — | — | | Veber | — | — | — | — | ≤10 | ≤140 | — | | Oprea (lead) | 200-350 | -2 to 4 | — | — | ≤7 | — | Rings ≤4 | | Leadlike Soft | 250-450 | -3 to 4 | — | — | ≤10 | — | — | | Leadlike Strict | 200-350 | -2 to 3.5 | — | — | ≤7 | — | Rings 1-3 | | CNS | ≤450 | -1 to 5 | ≤2 | — | — | ≤90 | — | | REOS | 200-500 | -5 to 5 | 0-5 | 0-10 | — | — | — | | Ro3 (fragment) | ≤300 | ≤3 | ≤3 | ≤3 | ≤3 | ≤60 | — | | Golden Triangle | 200-50*LogP+400 | -2 to 5 | — | — | — | — | — | | Rule of Drug | Ro5 + Veber + no PAINS | | | | | | |

Filter Selection by Discovery Stage

| Stage | Recommended Filters | Rationale | |-------|-------------------|-----------| | Initial screening | Ro5, PAINS, Common Alerts | Broad triage, remove obvious liabilities | | Hit-to-lead | Oprea or Leadlike Soft, NIBR, Lilly | Lead-like space, industrial filters | | Lead optimization | Rule of Drug, Leadlike Strict, Complexity | Strict drug-likeness + synthetic feasibility | | CNS targets | Rule of CNS, TPSA ≤90, HBD ≤2 | BBB permeability requirements | | Fragment-based | Ro3, low complexity (≤250) | Fragments have "room to grow" |

Rules Are Guidelines, Not Absolutes

~10% of marketed drugs violate Ro5. Exceptions are common for natural products, antibiotics, PROTACs, and prodrugs. Always combine rule-based filtering with domain expertise and target-class knowledge. Different modalities (oral, IV, topical) and target classes (kinases, GPCRs, ion channels) have distinct optimal property spaces.

Common Workflows

Workflow 1: Compound Library Triage

Full pipeline from raw SMILES to filtered drug-like candidates.

import pandas as pd
import datamol as dm
import medchem as mc

# Load compound library
df = pd.read_csv("compounds.csv")
mols = [dm.to_mol(s) for s in df["smiles"]]
print(f"Input: {len(mols)} molecules")

# Step 1: Drug-likeness rules
rfilter = mc.rules.RuleFilters(rule_list=["rule_of_five", "rule_of_veber"])
rule_results = rfilter(mols=mols, n_jobs=-1, progress=True)

# Step 2: Structural alerts
alert_filter = mc.structural.CommonAlertsFilters()
alert_results = alert_filter(mols=mols, n_jobs=-1, progress=True)

# Step 3: Combine results
df["passes_rules"] = [all(r.values()) for r in rule_results]
df["has_alerts"] = [r["has_alerts"] for r in alert_results]
df["drug_like"] = df["passes_rules"] & ~df["has_alerts"]

filtered = df[df["drug_like"]]
print(f"Output: {len(filtered)} drug-like molecules ({len(filtered)/len(df)*100:.1f}%)")
filtered.to_csv("filtered_compounds.csv", index=False)

Workflow 2: Lead Optimization Filter Cascade

Apply progressively stricter filters with detailed reporting.

import datamol as dm
import medchem as mc
import pandas as pd

mols = [dm.to_mol(s) for s in smiles_list]

# Cascade: rules → structural alerts → complexity → Lilly demerits
filters = [
    ("Leadlike Strict", mc.rules.RuleFilters(rule_list=["rule_of_leadlike_strict"])),
    ("NIBR Alerts", mc.structural.NIBRFilters()),
    ("Complexity ≤400", mc.complexity.ComplexityFilter(max_complexity=400)),
    ("Lilly Demerits", mc.structural.LillyDemeritsFilters()),
]

surviving = list(range(len(mols)))
for name, filt in filters:
    results = filt(mols=[mols[i] for i in surviving], n_jobs=-1)
    # Handle different result formats
    if isinstance(results[0], dict):
        passed = [i for i, r in zip(surviving, results)
                  if r.get("passes", not r.get("has_alerts", True))]
    else:
        passed = [i for i, r in zip(surviving, results) if r]
    print(f"{name}: {len(surviving)} → {len(passed)}")
    surviving = passed

print(f"Final candidates: {len(surviving)} / {len(mols)}")

Key Parameters

| Parameter | Module | Default | Range | Effect | |-----------|--------|---------|-------|--------| | rule_list | RuleFilters | — | See rules table | Which drug-likeness rules to apply | | n_jobs | All filters | 1 | -1 to N | Parallel workers (-1 = all co

Truncated for display — read the full file on GitHub.

Related Skills

View on GitHub
GitHub Stars367
CategoryDevelopment
Updated2mo ago
Forks36

Languages

Python

Trust signals

88/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

1 medium