Skip to content
antibacterial .ai

Compound versus strain screening

MIC prediction software: predict minimum inhibitory concentration in silico, before plate time

Paste a structure, pick a strain panel, and get a predicted MIC band per organism with the resistance mechanism named. The point is to decide which compounds are worth a plate, not to replace the plate.

Plans from $149 a month

Run a screen

of 3 left

Strain panel

No account needed. 3 screens per session.

Sample output Ciprofloxacin Fluoroquinolone
Worked example, replaced when you run
Strain MIC (µg/mL) Call Resistance risk Conf.
S. aureus (MRSA) 0.5->32 R Resistant target mutation, grlA and gyrA Fluoroquinolone resistance is widespread in methicillin resistant lineages.
E. coli <=0.015-0.06 S Susceptible efflux, AcrAB-TolC in resistant isolates Potent against wild type; qnr carriage and gyrA changes shift the band sharply.
K. pneumoniae (CRE) 0.5->64 R Resistant gyrA mutation with plasmid qnr Carbapenemase producing isolates almost always carry quinolone resistance too.
A. baumannii 8->64 R Resistant efflux, AdeABC Constitutive efflux plus target changes leave little room at achievable exposure.
P. aeruginosa 0.25-2 I Intermediate efflux, MexAB-OprM Borderline: active on many isolates, lost quickly once efflux is derepressed.
E. faecium (VRE) 4->32 R Resistant target mutation, parC Enterococci are intrinsically poor fluoroquinolone targets.

Why

Ciprofloxacin is a well characterised fluoroquinolone, so the Gram-negative bands are strongly supported by published activity data. The deciding factor across this panel is not target affinity but exposure: efflux in P. aeruginosa and A. baumannii, and acquired target mutation everywhere resistance is already common. Against a modern ESKAPE panel it reads as a Gram-negative agent with two reliable losses.

Closest published analogs

  • Levofloxacin Broadly similar Gram-negative bands, better Gram-positive coverage
  • Delafloxacin Retains activity against many ciprofloxacin resistant staphylococci
Strain MIC (µg/mL) Call Resistance risk Conf.

Why

Closest published analogs

Want the rest of your series? Batch runs are on the paid tiers. screen left in this session. This session is used up.

Computational prediction from published literature. Research use only, not a lab measurement and not clinical guidance.

The short answer

What MIC prediction software actually does

MIC prediction software estimates the minimum inhibitory concentration of a compound against named bacterial strains from its structure and the published record, before any plate is booked. It returns a predicted MIC band rather than a single number, because broth microdilution itself only resolves to doubling dilutions. The purpose is triage: deciding which twelve of four hundred analogs are worth assay time this month.

Antibacterial does this per strain rather than per target. You paste a SMILES string, an InChI or a compound name, pick a panel, and get a strain by read-out matrix: predicted band, an S, I or R call, the resistance mechanism the model expects, the closest published analogs with their measured activity, and how much support each row has. Low confidence is reported as low confidence.

At a glance

Input
SMILES, InChI or a compound name
Output
Predicted MIC band per strain, S/I/R call, mechanism
Panels
ESKAPE, Gram-negative, Gram-positive or custom
Grounding
Published MIC records for the closest analogs

What you get back

Five columns, and one of them is the confidence

01

Predicted MIC band

A range in µg/mL per organism, not a false-precision single value. Doubling dilutions mean the bench answer is plus or minus one step anyway, so a band is the honest unit.

02

S, I or R call

Read against the published breakpoint for that species where one exists. Where no breakpoint exists, the row says so rather than inventing a category.

03

Resistance mechanism

The mechanism the model expects to defeat the compound, named: beta-lactamase hydrolysis, target modification, RND efflux, or failure to cross the outer membrane.

04

Closest published analogs

The structural neighbours the call rests on, with their measured activity and the source. You can check the reasoning instead of trusting a score.

05

Confidence level

How well populated that neighbourhood is. A novel scaffold gets low confidence and a wide band, which is the correct answer, not a failure of the tool.

06

Export and API

CSV on every plan, SDF and PDF reports on Discovery and above, with API access for teams that want the read-out inside their own pipeline.

Those same fields come back as JSON from the antibacterial screening API, so a registry or a nightly triage job can store the matrix without anyone opening a browser. A whole deck goes the same way in one file through compound library screening.

Honest comparison

How in silico MIC prediction compares to the other ways of getting the answer

Each of these answers a different question. The table is about sequencing them correctly, not about claiming a prediction beats a measurement.

Approach Turnaround What it answers Where it falls short
In silico MIC prediction Seconds Which compounds in a series are worth a plate, and why the rest are not It is an estimate. It never becomes the number you publish or file
Broth microdilution in house Days to weeks The measured MIC against your isolates, the ground truth Plate time, media and compound are finite, so the queue sets the pace
Contract research organization panel Weeks, plus contracting A measured panel without building the capability yourself Quoted per project, and the quote arrives before you know which compounds deserve it
Free academic screening schemes Months, when accepted A measured primary screen at no cash cost to an academic group Queues, eligibility rules, and results that enter an open database after an embargo
Manual literature and ChEMBL search An afternoon per compound What has already been reported for close analogs It does not scale past a handful of compounds and it produces no ranking

The sequence that works is prediction first, then a smaller confirmatory panel. Prediction is cheap and wrong sometimes; a plate is expensive and right. Spending the expensive thing on compounds the cheap thing already flagged as dead is the waste worth removing. If you want the bench side in detail, the method behind a measured MIC is written up separately.

Who it is for

Four people who ask this question for a living

Medicinal chemist on an antibacterial program

You have four hundred analogs and plate capacity for twelve. The read-out ranks them and gives you a mechanism-level reason for every compound you leave out, which is what makes the decision defensible in a project meeting.

Assay lead at a discovery CRO

You quote panels for clients and absorb the cost of submissions that were never going to work. Pre-triaging incoming compounds lets you scope a realistic panel and set expectations before the plates are committed.

Academic AMR lab principal investigator

Assay time is the scarce resource and the grant is finite. A literature-grounded ranking is the evidence that justifies which natural-product or peptide leads get the limited bench slots.

Head of discovery holding the budget

You own the go or no-go on a scaffold. Killing a dead-end series at week two instead of month five is the single largest line item you control, and the mechanism call is what makes an early kill survive scrutiny.

Strain panels

The unit of work is compound times strain

A general discovery suite scores a compound against a target. An antibacterial program needs the answer per organism, because the same molecule can be excellent against MRSA and completely inert against Gram-negative bacteria for reasons that have nothing to do with the target.

ESKAPE

Six pathogens behind most resistant hospital infections.

  • S. aureus (MRSA)
  • E. coli
  • K. pneumoniae (CRE)
  • A. baumannii
  • P. aeruginosa
  • E. faecium (VRE)

Gram-negative

Permeability and efflux decide almost everything here.

  • E. coli
  • K. pneumoniae (CRE)
  • A. baumannii
  • P. aeruginosa
  • E. cloacae
  • S. Typhimurium

Gram-positive

Target modification and acquired operons dominate.

  • S. aureus (MRSA)
  • S. aureus (MSSA)
  • E. faecium (VRE)
  • E. faecalis
  • S. pneumoniae
  • S. epidermidis

Custom panels and your own clinical isolates are available on the Discovery plan and above. Peptide leads run through the same panels, with the sequence-specific read-out described on the antimicrobial peptide prediction tool page.

Questions people ask before buying

MIC prediction, answered plainly

Can MIC be predicted in silico?

A MIC band can be predicted in silico, a single exact number cannot. Broth microdilution reads in doubling dilutions, so the honest unit of prediction is a band such as 1 to 4 µg/mL. Antibacterial reports that band per strain, with a confidence level and the published analogs the call rests on, and says so when support is thin.

How accurate is MIC prediction software?

Accuracy depends almost entirely on how well populated the published neighbourhood is. A compound close to well characterised analogs gets a tight band and high confidence; a genuinely novel scaffold gets a wide band and low confidence. Any tool that returns the same confident number for both is hiding the thing you most need to know.

What is the difference between MIC prediction and a MIC assay?

A MIC assay measures activity on a plate against a real isolate and is the ground truth. MIC prediction estimates that result from structure and published data before any plate is booked. Prediction is for triage, deciding which compounds earn assay time. It does not replace the assay and no regulator accepts it as a substitute.

Does MIC prediction work for Gram-negative bacteria?

It works, and this is where it pays off most. Most compounds fail Gram-negatives on uptake and efflux rather than on target affinity, and those failures are the ones that waste plate time. Flagging a likely efflux substrate or a molecule that will not cross the outer membrane kills a dead end weeks earlier.

Can I screen antimicrobial peptides as well as small molecules?

Yes. Peptide leads can be submitted as a sequence or a structure and run against the same panels. The read-out flags the usual peptide liabilities the literature reports for close relatives, including haemolysis risk and protease stability, alongside the predicted band per organism.

Do my compounds stay confidential?

Structures you submit stay yours and are never used to train models. That matters for a program with unpublished chemistry, and it is the practical reason a biotech cannot simply route everything through a free academic screening scheme, where data is shared into an open database after an embargo period.

Pricing

What MIC prediction software costs

Full plan comparison

The number is on the page, which is not the norm in this market. Almost every vendor selling computational work into drug discovery quotes on request, and the published ranges that do exist start in five figures a year. How the wider category discloses, vendor by vendor, is set out on the drug discovery software cost page. A file of structures rather than one compound goes through compound library screening software.

Yearly billing is two months cheaper, charged once a year.

Lab

$124 /mo

$1,488 charged once a year

250 screens a month against the standard panels, with CSV export.

Get started

Academic AMR lab, 1 to 2 chemists

Discovery

Most popular

$416 /mo

$4,992 charged once a year

Custom panels, batch runs of a full series, API access and roles.

Get started

Biotech antibacterial program

Program

$1,241 /mo

$14,892 charged once a year

Library-scale runs, your own isolates, SSO, audit log and an SLA.

Get started

Multi-program biotech or discovery CRO

Enterprise

Talk to sales

Custom terms, invoiced

Private deployment, your internal isolate library, a named contact.

Pharma, institute or consortium

Screen a compound before you book the plate

Three screens per session, no account needed. Paste a structure, pick a panel, read the matrix.

Run a screen first

What you are agreeing to

  • Your compounds stay yours
  • Never used for model training
  • Deletable on request
  • No card needed to create your account

Related reading

The reference pages behind the calls the screen makes.

Screen your own compound.
No card required.