How to Write a Molecular Docking and MD Results Section

The results section of a docking and MD study reports what you observed, not what it means. Give binding affinities in a table with the redocking RMSD control beside them, one figure per convergence claim, captions that stand alone, and every number with its unit and uncertainty. Interpretation belongs in the discussion.

Most computational thesis chapters lose marks in the same place. The docking ran, the simulation finished, the analysis scripts produced their output, and then the student pastes nine lines of AutoDock Vina output and four unlabelled plots into a chapter and calls it results. The science was fine. The reporting was not. This guide covers the reporting craft only, written by the StemSkills Lab team from 10+ years in structural bioinformatics, drug design and multiscale molecular modeling.

If you have not yet decided what your numbers mean, read our guide on how to interpret molecular docking results first. This article assumes the interpretation is done and the writing is what remains.

What belongs in the results section and what belongs in the discussion?

This is the single most common objection in a viva, and the rule is not a matter of taste. The ICMJE Recommendations, the style guidance most life-science journals defer to, ask authors to present results “in logical sequence in the text, tables, and figures”, giving the most important findings first. For the discussion, the same document says explicitly not to repeat in detail data already given in the results.

Applied to a docking and MD chapter, that splits cleanly:

• Results: the affinity values, the poses you obtained, the residues in contact, the RMSD plateau, the RMSF peaks, the hydrogen bond occupancies, the MM-PBSA numbers with their error. Observations.
• Discussion: why compound 4 outscored the reference, what the flexible loop implies for selectivity, how your findings compare with the published inhibitor series, what the limitations of an implicit-solvent end-point method are. Explanations.

A quick test before you keep a sentence in results: if it contains the word because, suggests, indicates, may explain or which is consistent with, it is a discussion sentence. Move it. The one exception is a purely operational because (“frames before 20 ns were excluded because the backbone RMSD had not levelled off”), and even that usually belongs in the methods.

What does a molecular docking results table need?

One table, one row per ligand, top-ranked pose only. A table that reproduces every pose Vina printed is unreadable and tells the examiner you did not choose. Put the full pose list in supplementary material if you want it on record.

Two rules about size. Tabulate every ligand you screened only if there are fewer than about fifteen; beyond that, tabulate the top hits plus the control and send the rest to supplementary with a sentence saying how many were screened and what the cut-off was. And never present a table without the control row. A column of negative numbers with nothing to compare against is not a result.

How do I report binding affinity so the numbers are believable?

Report the score and the validation of the protocol that produced it, side by side. The validation is a redocking run: take the ligand out of its crystal complex, dock it back with the settings you used for everything else, and measure the RMSD between the redocked pose and the crystallographic pose.

The threshold is not folklore. In the original AutoDock Vina paper, Trott and Olson (2010) state that an RMSD cut-off of 2 Å is often used as the criterion of a correct bound-structure prediction, and they evaluate accuracy on a test set of 190 complexes using exactly that cut-off. So one sentence at the top of your results carries a lot of weight:

Redocking of the co-crystallised ligand reproduced the experimental pose with an RMSD of 1.2 Å, below the 2 Å criterion, so the protocol was applied unchanged to the screening set.

Fill in your own number. If your redocking RMSD is above 2 Å, report it honestly and say what you changed, because the alternative is an examiner finding it for you.

Three more habits that make affinity reporting defensible:

• Quote the program’s own precision. The AutoDock Vina basic docking tutorial prints affinities to two decimal places under the headings affinity (kcal/mol) and dist from best mode: rmsd l.b., rmsd u.b.. In that tutorial’s own imatinib and Abl kinase example (PDB 1IEP), the best mode comes out at -14.72 kcal/mol with the AutoDock4 scoring function. Copy that precision. Do not add decimals the program did not give you, and do not round to whole numbers.
• Call it what it is. A Vina score is a predicted binding affinity from an empirical scoring function. It is not a measured free energy of binding. Write “predicted binding affinity” in results and save the caveat for the discussion.
• Do not make claims a scoring function cannot support. A gap of 0.2 kcal/mol between two ligands is inside the noise of the method. Report the rank order, report the values, and resist writing that one compound binds more strongly.

Your methods section has to match this. If you have not written it yet, our guide to the methods section of a molecular docking study covers the settings you must declare, and the GROMACS methods section guide does the same for the simulation.

Which MD plots earn a figure and which belong in supplementary?

The working rule: a plot earns a place in the main text when a sentence in your results depends on it. If no sentence would change were the plot deleted, it is supplementary. Six analyses come out of a standard protein-ligand run, and they do not have equal claim on your page budget.

Two presentation rules apply to all of them. Mark the equilibration you discarded, either by shading the region or by starting the analysis after it and saying so in the caption. And overlay the apo and holo traces on one axis rather than printing two separate plots, because the comparison is the result and a reader should not have to hold one figure in memory while looking at another. Our walkthroughs on RMSD and RMSF analysis and on plotting GROMACS xvg files in Python cover the mechanics.

How do I write a figure caption that stands alone?

Assume the caption will be read by someone who has not read your text, because in peer review and in a viva that is what happens. A complete caption answers five questions: what is plotted, for what system, over what time, computed how, and what the reader should notice.

Weak: Figure 3. RMSD plot.

Complete: Figure 3. Backbone RMSD of the apo protein (grey) and the protein-ligand complex (blue) over 100 ns of unrestrained MD, computed with gmx rms after least-squares fitting to the energy-minimised starting structure. Both systems level off within the first 15 ns, which was excluded from all subsequent analysis.

The last sentence is the one students leave out, and it is the one that does the work. Note that it describes what is visible, not what it means. A caption saying the complex is more stable has crossed into the discussion. Keep the styling consistent too: same fonts, same axis conventions and same colour assignment across every figure, which our guide to publication-quality figures in PyMOL covers for the structural images.

How do I report MM-PBSA binding free energies correctly?

An MM-PBSA number without an error estimate is not reportable. These are averages over frames, and the spread between frames is often large, so give the mean with its standard error of the mean and state how many frames it came from and which window of the trajectory they were drawn from.

A defensible sentence looks like this:

The binding free energy calculated with gmx_MMPBSA over 500 frames sampled from the final 50 ns was -38.4 ± 2.1 kcal/mol for compound 4 and -29.7 ± 2.6 kcal/mol for the reference inhibitor.

Present the decomposed terms (van der Waals, electrostatic, polar and non-polar solvation) in a table rather than in prose, because five numbers per ligand read as noise in a sentence and as a pattern in a table. Report entropy only if you calculated it, and say which method you used. The gmx_MMPBSA documentation is the reference for what each output term contains, and our tutorial on binding free energy with MM-PBSA in GROMACS covers the run itself.

One thing to keep out of the results: the observation that end-point methods overestimate absolute binding energies. That is true, it matters, and it belongs in the discussion or the limitations.

Which units, symbols and decimal places are correct?

Mixed units are the cheapest mark to lose and the easiest to fix. The conversion is exact: 1 kcal = 4.184 kJ, and 1 Å = 0.1 nm.

• Energies: docking programs report kcal/mol, GROMACS reports kJ/mol. Pick one for the whole chapter, convert everything to it, and state the conversion once in methods. Do not present a docking table in kcal/mol and an MM-PBSA table in kJ/mol three pages later.
• Distances: the PDB and docking output use Ångström, GROMACS uses nanometres. Same rule. If you convert a GROMACS RMSD to Ångström for consistency with your docking section, say so in the axis label, not only in your head.
• Decimal places: two for energies in kcal/mol, two for RMSD in Ångström, one for percentages and occupancies. More digits imply a precision the method does not have.
• Time: ns for trajectory length, ps for output frequency, fs for the integration step. State the production length, not the total including equilibration.
• Residues: three-letter code plus number plus chain, as in Thr315 (chain A), and say whether numbering follows the PDB entry or your own renumbering after modelling missing loops.

What does a finished results paragraph actually look like?

Here is the shape, with the parts you replace in square brackets. Structure every paragraph as claim, then evidence, then pointer to the table or figure.

All [N] ligands were docked into the [site name] of [protein, PDB ID, chain]. Predicted binding affinities ranged from [x] to [y] kcal/mol (Table 1). [Compound ID] ranked highest at [value] kcal/mol, compared with [value] kcal/mol for the reference ligand [name] run under identical settings. The top-ranked pose of [compound ID] occupied the same subpocket as the co-crystallised ligand and formed [n] hydrogen bonds, with [residue list] (Figure 2). Redocking of [reference ligand] reproduced the crystallographic pose to within [value] Å.

Every sentence is an observation. Nothing explains, ranks by merit, or speculates. The discussion paragraph that follows in your chapter is where the ranking becomes an argument. If you are still choosing which compound to carry forward into MD, our guide on reading docking output covers the decision, and the computational biology skills roadmap shows where writing sits in the wider sequence of skills.

Troubleshooting: real viva objections and the fix

• “Why is this sentence in your results?” You explained a mechanism. Cut the clause and move it to the discussion. Search your chapter for because, suggests and indicates and check each one.
• “How do you know the docking protocol works?” No redocking control. Add the redocking RMSD sentence and the control row in Table 1.
• “Which part of the trajectory is this?” The equilibration is unmarked. State the discarded window in the caption and in methods.
• “Your affinities differ by 0.3 kcal/mol. Is that meaningful?” Do not defend the gap. Report the rank order and note that the values fall within the resolution of the scoring function.
• “Where is the error on this energy?” An MM-PBSA mean with no standard error and no frame count. Recompute and report both.
• “Why nine rows for one ligand?” You tabulated every pose the program printed. Keep the top-ranked pose per ligand and move the rest to supplementary.
• “Are these the same units?” kcal/mol in the docking table, kJ/mol in the MM-PBSA table. Convert the whole chapter to one.
• “Which chain, and whose numbering?” Add the chain identifier to every residue and state the numbering convention once.
• “Your figure has no axis label.” Check every axis for quantity and unit. A bare number on an axis is not a label.

Frequently asked questions

How long should the results section of an MSc dissertation chapter be?

Your department handbook sets the limit, so check it first. As a working guide, budget the length from your figures and tables rather than a word count: one table or figure per claim, and text that adds what the visual cannot show. Repeating the numbers already in a table is padding, and the ICMJE guidance asks you not to do it.

Should I report all docked poses or only the best one?

The top-ranked pose per ligand in the main table. The full pose list, with the RMSD columns the program prints, goes to supplementary. An exception applies when two poses score within the noise and occupy different subpockets, which is itself an observation worth one sentence.

Can I call the AutoDock Vina score a binding free energy?

No. It is a predicted binding affinity from an empirical scoring function fitted to experimental data. Write “predicted binding affinity (kcal/mol)” in your table heading. Treating a docking score as a measured thermodynamic quantity is a claim the method does not support.

Do results come before or after figures in the chapter?

Follow the template your university or target journal specifies. Whichever applies, every figure and table must be cited in the text in the order it appears, and each must be numbered and captioned so it can be read on its own.

Can I include statistics in the results section?

Yes, provided the test was named in your methods. Report the test, the statistic, the sample size and the p-value, and describe the comparison without interpreting its significance for your hypothesis. That interpretation is a discussion sentence.

What if my results do not support my hypothesis?

Report them exactly as they are. A negative or inconclusive result is a result, and examiners mark the rigour of the work, not the direction of the outcome. Adjusting a protocol until the numbers agree with the hypothesis, without saying so, is the one error that cannot be fixed at the writing stage.

Where this fits in the wider workflow

Writing is the last step of a chain that starts with target selection and runs through preparation, docking, simulation and analysis. Each earlier step constrains what you are allowed to claim at the end, which is why the results section is easiest to write when the methods were recorded as you went. Our molecular docking pillar guide maps that full sequence, and the individual tutorials cover each stage in detail.