Caput Designdraft
EN – EnglishTR – Türkçe

Approach

Knowing what a score means

The value of a design protocol depends on knowing the scale on which its numbers should be read. So we define every metric we deliver, state how it should be interpreted and say plainly what it does not express.

The metrics and how to read them

Interface confidence metrics
What it showsIndicate how consistently the model builds the interface. Used to rank candidates against each other.
What it does not showNot an affinity. A high value does not mean the design will bind.
Binding energy estimate
What it showsAn estimate computed from the structure, used to see which candidate stands out.
What it does not showNot a measured binding free energy; not directly comparable with an experimental value.
Buried surface area and packing quality
What it showsDescribe the size of the interface and its geometric fit.
What it does not showDo not by themselves determine binding strength.
Residue-level contact list
What it showsGives which residues face each other, and at what distance, within the model. The cutoff and numbering are stated in the report.
What it does not showNot an experimentally verified interaction list.
Epitope conservation / selectivity table
What it showsShows how conserved the surface is across the paralogs you name.
What it does not showNot measured selectivity; a sequence and structure based prediction.
MD stability summary
What it showsSummarises how stable the complex geometry stays over the simulated window.
What it does not showNot a measured in-cell lifetime; the duration and conditions are stated in the report.
Docking score (small molecule)
What it showsUsed to rank poses against each other. A retrospective discrimination test for the target class is provided alongside.
What it does not showNot a binding free energy.

Calibration and validation

For a protocol’s scores to mean anything they have to be tested against known results. We do not publish a numerical calibration result on this page yet: we would rather not put the relevant claims live before their evidence is approved. Once approved, which metric came how close on which reference will be written here, with the source.

Claims not published at present

The claims listed below appear in the source draft but are not used on the site because they await evidence and approval. This list is visible only in the draft build.

C01 · C02 · C03 · C04 · C05 · C06 · C07 · C08 · C09 · C10

Deliverables

The files you receive at the end of a project

For every file: what it is, what it does not show, and which projects produce it. This list is the content of the delivery package, not a sample report.

  • Candidate sequences

    Every project
    candidates.fasta
    What it contains
    Ranked candidate sequences, each with an identifier
    What it does not show
    The ranking follows computational criteria; it does not guarantee the experimental order.
    What does the file look like?Format example
    >BARNASE_1BRS_A | gosterim dizisi, tasarlanmis aday DEGILDIR
AQVINTFDGVADYLQTYHKLPDNYITKSEAQALGWVASKGNLADVAPGKSIGGDIFSNRE
GKLPGKSGRTWREADINYTSGFRNSDRILYSSDWLIYKTTDHYQTFTKIR

    In a real delivery each record is a designed candidate sequence carrying its rank. The sequence above is taken from the barnase chain in the 3D demonstration, purely to show the format.

  • Complex models

    Every project
    complexes/*.pdb
    What it contains
    Predicted target–binder complex for every candidate
    What it does not show
    These are predicted structures, not experimentally solved ones.
    What does the file look like?Format example
    complexes/
  candidate_001.pdb
  candidate_002.pdb
  ...
  candidate_NNN.pdb

    One structure file per candidate. File names match the candidate identifiers.

  • Evaluation table

    Every project
    metrics.csv
    What it contains
    Per-candidate interface confidence scores, binding energy estimate, buried surface area, packing quality
    What it does not show
    Interface confidence metrics and energy estimates are selection criteria, not affinity measurements.
    What does the file look like?Format example
    candidate_id,interface_confidence,binding_energy_estimate_kcal_mol,buried_surface_area_A2,packing_quality
<kimlik>,<0-1>,<kcal/mol>,<A^2>,<0-1>

    Column structure only. No sample numbers: an invented metric value could be read as a real result.

  • Contact list

    Every project
    interface_contacts.csv
    What it contains
    Residue-level contact list in the target's own numbering
    What it does not show
    Contacts are read from the model; this is not an experimentally verified interaction list.
    What does the file look like?Real example
    target_chain,target_resi,target_resn,target_atom,partner_chain,partner_resi,partner_resn,partner_atom,min_distance_A
A,83,ARG,NH2,D,39,ASP,OD1,2.5
A,83,ARG,O,D,29,TYR,OH,2.65
A,102,HIS,NE2,D,39,ASP,OD2,2.81
A,60,GLU,OE2,D,34,LEU,N,2.84
A,59,ARG,N,D,35,ASP,OD1,2.88

    These rows are real: computed from PDB 1BRS at a 4.5 Å cutoff. The full file has 43 rows.

  • Selectivity table

    Scope dependent
    selectivity.csv
    What it contains
    Epitope conservation table against the paralogs you name
    What it does not show
    A sequence and structure based prediction, not measured selectivity.
    What does the file look like?Format example
    candidate_id,paralog,epitope_identity,conserved_positions,differing_positions
<kimlik>,<UniProt>,<0-1>,<liste>,<liste>

    Column structure. You decide which paralogs are compared.

  • Stability summary

    Scope dependent
    md_summary.pdf
    What it contains
    RMSD/RMSF and contact persistence for the selected candidates
    What it does not show
    Scope dependent: not produced in every package or for every candidate.
    What does the file look like?Format example
    1. Simulasyon kurulumu (sure, kosullar, kuvvet alani)
2. RMSD / RMSF egrileri
3. Temas surekliligi tablosu
4. Yorum ve sinirlar

    Report sections. Duration and conditions are stated per project.

  • Method report

    Every project
    report.pdf
    What it contains
    Method summary, ranking rationale and methodological limits
    What does the file look like?Format example
    1. Hedef ve kapsam
2. Yontem ozeti (kullanilan adimlar)
3. Siralama gerekcesi
4. Aday bazinda bulgular
5. Metodolojik sinirlar

    Report sections. The “methodological limits” section appears in every report.

  • Benchmark report

    Scope dependent
    benchmark.pdf
    What it contains
    Small-molecule work: retrospective discrimination result for the target class
    What it does not show
    Specific to the small-molecule arm; not a mandatory part of the protein package.
    What does the file look like?Format example
    1. Hedef sinifi ve secilen referans seti
2. Bilinen aktif / yem bilesik ayirici testi
3. Sonuc ve yorum

    Report sections.

Process

How the work proceeds, step by step

Because timing and scope differ per project, no calendar is stated here; both are put in writing in the scoping call.

  1. 01

    Scoping call

    The target, the epitope of interest, the selectivity requirement and the delivery expectation are agreed.

  2. 02

    Target preparation

    Structure selection, preparation and definition of the region to be worked on.

  3. 03

    Design and filtering

    Candidate generation, filtering and ranking on interface confidence metrics.

  4. 04

    Interface and selectivity

    Contact analysis and selectivity assessment against the paralogs you name.

  5. 05

    Stability

    Molecular dynamics on the leading candidates to examine complex stability.

  6. 06

    Report and delivery

    Files, the method report and the methodological limits are delivered together.

  7. +

    Experimental validationOptional scope

    Optional: binding measurement at an external laboratory, compared against the computational ranking.

Terms used on this page

New to the field? Start here. The explanations are deliberately short.

interface
The surface where two proteins touch. Whether a design works shows up largely here.
affinity
How tightly two molecules hold on to each other. It is measured in the lab, not computed.
computational
Work done on a computer. It is not an experiment; it shows which candidate is worth testing.
residue
A single amino acid in the protein chain — one link in the chain.
epitope
The specific patch on the target’s surface where you want the binding to happen.
paralog
A related protein that resembles the target. You usually want the design not to bind these by mistake.
selectivity
How well a design binds the target while leaving similar proteins alone.
complex
The structure formed by two or more molecules bound together.
molecular dynamics
Simulating how a structure moves over time — it shows whether a complex holds together.
RMSD / RMSF
Measures of how far a structure drifts (RMSD) and which parts wobble most (RMSF) during a simulation.
docking
Computing how a small molecule might sit in a pocket on the target.
structure
The three-dimensional shape of a protein — a map of where each atom sits.
PDB
The public database of experimentally solved protein structures.
UniProt
The public database of protein sequences and their basic annotations.

Describe your target briefly and we will settle the scope together.

The first message never asks for confidential sequences, unpublished structures or sensitive files. Technical detail follows an appropriate confidentiality process.

Let us discuss your project