Phage T4, built in ChimeraX

Inspired by Tom Goddard's 2022 ChimeraX animation of phage T4 tail contraction (RBVI, UCSF). This version uses T4's own tail, head and baseplate structures released since then. It adds simulated host membranes and close-ups, with a build you can repeat by asking Pellaeon.

Tail contraction, slowed down. The real motion takes under a millisecond.

The film

The player loads from youtube-nocookie.com only when you press play. Watch on YouTube.

Phone cut, 9:16: watch.

85 seconds, rendered in ChimeraX by a movie script from the same models: 5,094 frames at 3840 × 2160, about 9 hours on one GPU. The proteins use deposited models; the membranes are simulations. See what is schematic or not shown.

Close-ups

Twelve views of the same models. Select a numbered dot to name a part. Drag a slider to compare two states; ⤢ opens a larger view.

A peach ring of gp3 with a dark channel in its centre, green gp15 crescents around it, the head lattice behind — **Looking toward the head** Looking from the tail toward the head. Six-fold rings of gp3 and gp15 sit below the twelve-fold portal. PDB 9MKB, 7VS5

A red ribbon β-helix pointing right, ending in a thin tip, purple baseplate out of focus on the left — **The spike tip** The gp5 β-helix ends in gp5.4, the small protein at the spike tip. PDB 9MKB

A vertical tube cut lengthwise: yellow cut faces, a pink and white lining, blue sheath behind — **Inside the tube** The tail tube, cut lengthwise. ChimeraX colors the lining by computed electrostatic potential: red negative, blue positive (-10 to +10 kcal/mol·e). PDB 9MKB

The baseplate seen from below, a compact hexagon before and an open six-pointed star after the trigger: Before the trigger, 9MKB — **Hexagon to star** The baseplate from below: a hexagon about 52 nm across opens to a star about 61 nm across. A hint of Demogorgon. The gp12 fibers stay folded; their swing is schematic. PDB 9MKB, 9X3Q, 9X3R, 5IV7

The baseplate seen from below, a compact hexagon before and an open six-pointed star after the trigger: After, 9X3Q + 9X3R — **Hexagon to star** The baseplate from below: a hexagon about 52 nm across opens to a star about 61 nm across. A hint of Demogorgon. The gp12 fibers stay folded; their swing is schematic. PDB 9MKB, 9X3Q, 9X3R, 5IV7

A white and lilac β-barrel trimer standing in a membrane of blue and gray spheres — **Where the fibers bind** OmpC, a receptor for T4 on E. coli K-12, in a simulated outer membrane. The cut passes through the trimer. PDB 2J1N; membrane: Gutishvili 2025

The capsid surface in low light, rounded hexagonal bumps outlined in pink — **The head lattice** gp23* hexamers form the head lattice, with Soc between them. PDB 7VS5

A slab of the sheath as a gray density surface and as blue atoms: Cryo-EM map, EMD-48324 — **Map and model** The same section of tail as cryo-EM density (contour 0.012) and as the atomic model built into it. 84 % of the atoms of one gp18 subunit lie inside the contour. EMD-48324, PDB 9MKB

A slab of the sheath as a gray density surface and as blue atoms: Atomic model, 9MKB — **Map and model** The same section of tail as cryo-EM density (contour 0.012) and as the atomic model built into it. 84 % of the atoms of one gp18 subunit lie inside the contour. EMD-48324, PDB 9MKB

Ribbon model of a spike hanging under purple baseplate proteins — **The spike** gp27, three gp5 chains and the gp5.4 tip: a spike about 19 nm long under the baseplate. PDB 9MKB

One sheath subunit as a blue ribbon with a green arm and a peach arm that swings up in the contracted state: Extended, 9MKB — **A sheath subunit in two states** The same gp18 subunit from the extended and contracted sheath, aligned on its core. Its core matches within 1.6 Å over 611 Cα pairs; the two arms that grip neighboring subunits move. PDB 9MKB, 9X3P

One sheath subunit as a blue ribbon with a green arm and a peach arm that swings up in the contracted state: Contracted, 9X3P — **A sheath subunit in two states** The same gp18 subunit from the extended and contracted sheath, aligned on its core. Its core matches within 1.6 Å over 611 Cα pairs; the two arms that grip neighboring subunits move. PDB 9MKB, 9X3P

A tunnel of pink and blue surface rings receding to a dark centre — **Along the tail tube** Looking along 50 nm of gp19 rings. Red marks negative and blue positive computed electrostatic potential. PDB 9MKB

Stick model of a short stretch of protein inside a translucent gray density — **Side chains in density** gp18 residues 200 to 232 in the cryo-EM map EMD-48324. Only density within 2.6 Å of the model is shown. EMD-48324, PDB 9MKB

A side cutaway of the neck: purple portal, blue and teal rings, green and peach rings below, the head shell above — **Inside the neck** The front half of the neck is removed to show where twelve-fold rings meet six-fold rings. PDB 9MKB, 7VS5

What you are looking at

The whole T4 virion from the side: a prolate head on the left, the neck, a striped sheath and the baseplate on the right — T4 assembled from 9MKB and 7VS5, shown from the side without perspective. The scale bar applies to this view. Published dimensions: Leiman et al. 2010; Fang et al. 2022. The model measures 243 nm from baseplate to head tip.

Infection in six steps

SearchLong tail fibers, 144 to 160 nm, bind OmpC or lipopolysaccharide (LPS). Binding by three fibers commits the phage to infection. not shown
TriggerThe baseplate opens from a 52 nm hexagon to a 61 nm star; in the model gp11 turns 95°. The short fibers unfold onto the LPS core. fiber swing schematic
ContractionA wave runs from the baseplate to the neck through all 23 sheath rings: rise 39.7 to 16.2 Å, twist 17.2 to 32.9° per ring. The sheath shortens from 925 to 420 Å, the head moves about 50 nm and the tube turns about 345°. In reality this takes under a millisecond.
Outer membraneThe tube pierces the outer membrane. About 16 nm of the spike, comprising gp5 and gp5.4, is lost on the way.
PeriplasmThe gp5 lysozyme digests peptidoglycan, the cell-wall mesh. The inner membrane, about 30 nm away, bulges about 16 nm to meet the tube tip and fuses with it. bulge not shown
Genome168,903 base pairs packed at about 500 mg/ml enter the cell at up to 3 to 4 kbp per second. not shown

Schematic or not shown

Long tail fibers and the fibritin whiskers: not shown. Deposited structures contain only short pieces.
Short tail fibers (gp12): no extended structure exists. Here they stay folded and swing as rigid bodies with their gp11 trimers.
Contraction wave: the sheath is interpolated between deposited end states (9MKB, 9X3P). The wave's shape and speed are modeled, not imaged.
Membranes: tiled snapshots of all-atom simulations, not experimental structures. LPS covers about 67 % of the outer leaflet; the tile seams are unrelaxed.
Inner-membrane bulge and DNA entry: not shown.
Time: stretched. The real contraction takes under a millisecond.

Build it with Pellaeon

Type these fourteen requests into Pellaeon in order. Each step shows the request, the commands it ran and the resulting view. By default, running scripts, deleting atoms and saving files wait for your OK.

Use ChimeraX 1.9 or newer. Allow about 6 GB of free memory and 64 MB of downloads (cost notes). Download t4_parts.py, t4_trigger.py and t4_wave.py from examples/phage-t4 (MIT license), and put them together in ~/phage-t4. Start with an empty ChimeraX session so the model numbers match the requests.

Tried on

One desktop PC: Ubuntu Linux, 32 CPU threads, about 120 GB of RAM and an NVIDIA RTX 4070 SUPER (12 GB). Software: UCSF ChimeraX 1.12 and Pellaeon 0.2.13.

The recorded run used GPT-6 Luna through a ChatGPT sign-in. Steps 5, 6, 8, 9 and 10 used GPT-6 Sol, a stronger model, because Luna did not handle them reliably.

With an earlier wording of these requests, local gemma4:12b on Ollama handled about half the steps, including opening structures, running scripts, saving and playing. It got the rest wrong, including wrong model numbers, chain IDs in the wrong case and selections that matched nothing.

The whole build peaks at about 6 GB of memory in the window. Drawing 5.3 million atoms is the heavy part. Different models, versions and machines will behave differently. Your mileage may vary.

Fourteen requests in 5 min 18 s. Model waits and animation playback are sped up and marked. One request where the model asked a question instead of acting is cut.

1. Open the tail

open 9mkb, the phage T4 tail with its neck and baseplate

open 9mkb

9MKB contains the mature virion's tail, neck and baseplate, plus the five capsomers around the portal: 1,533,268 atoms in 541 chains.

The 9MKB baseplate seen end-on in ChimeraX, the request and reply in the Pellaeon panel

2. Add the head as instances

open 7vs5, the T4 head, and show its assembly 1 as graphical instances, not copies (copies false, newModel false), then fit both models in the view

open 7vs5
sym #2 assembly 1 copies false newModel false
view

7VS5 contains one fifth of the head (781,038 atoms). Assembly 1 displays it five times: 3.9 million atoms on screen, with one copy in memory. Keep the head as graphical instances to keep memory low.

The head and the tail side by side, both still in their deposited positions

3. Put the tail on the head

put the tail on the head: run matchmaker #1/Ae,Ai,Am,BJ to #2/dt,du,dv,dw pairing ss computeSS false

matchmaker #1/Ae,Ai,Am,BJ to #2/dt,du,dv,dw pairing ss computeSS false

Both entries contain capsomers beside the portal. Matching four corresponding gp23 chains places the tail on the head. Three details matter:

Move the tail, not the head. Moving the instanced head would place only its first copy; moving the tail keeps all five together.
Give the chain IDs. With 541 chains in 9MKB, the model does not find them on its own.
computeSS false skips recalculating secondary structure for 2.3 million atoms: 11 s and 3 GB instead of 120 s and 46 GB.

One capsomer places the head within about 15 Å of a fit on all five, close enough at this scale.

The tail seated in the head, seen end-on from the baseplate

4. Name the parts

run the script ~/phage-t4/t4_parts.py

open ~/phage-t4/t4_parts.py

t4_parts.py gives groups of chains names such as head, neck, sheath and hub. Later requests can use these names instead of chain IDs.

The OK card for the script, and the reply listing the named parts

5. Remove the duplicate capsomers

delete capsid9mkb: those gp23 and Soc chains of #1 are already in the head model

delete #1 & capsid9mkb

The head model already supplies these parts. Remove the overlapping 30 gp23 and 10 Soc chains from 9MKB.

6. Color by part

color the named parts as they are, with the Catppuccin Mocha palette: color head #F5E0DC, soc #F5C2E7, neck #B4BEFE, sheath #89B4FA, tube #F9E2AF, inner #CBA6F7, outer #94E2D5, socket #FAB387, stf #A6E3A1, hub #F38BA8, other #BAC2DE

color head #F5E0DC
color soc #F5C2E7
color neck #B4BEFE
color sheath #89B4FA
color tube #F9E2AF
color inner #CBA6F7
color outer #94E2D5
color socket #FAB387
color stf #A6E3A1
color hub #F38BA8
color other #BAC2DE

The baseplate in teal and green, the head behind it in rose with pink Soc

7. Show the whole virion

show #1 and #2 as spheres, background #1E1E2E, lighting full, silhouettes on in #0E0F18, and show the whole virion from the side

show #1 atoms
show #2 atoms
style #1 sphere
style #2 sphere
set bgColor #1E1E2E
lighting full
graphics silhouettes true color #0E0F18
view

The sphere view shows the packed atoms. Shadows and outlines separate the parts.

The whole virion from the side as spheres with shadows and outlines

8. The baseplate from below

look at the baseplate from below, straight up the tail axis

turn y 180
view hub
view outer

9. The neck

turn back to a side view of the whole virion, then zoom in on the neck part, where the tail joins the head

view neck

Close-up of the neck between the head and the sheath

10. The cell-puncturing spike

hide everything except hub, show hub as cartoon and give me a side-view close-up of the needle

hide #1,2 target acs
cartoon hub
view hub
zoom 1.7

The hub as a cartoon: gp27, gp5 and the gp5.4 tip

11. Save an image

save this view with the save command as ~/phage-t4/needle.png, 1920 by 1080, supersample 3

save ~/phage-t4/needle.png width 1920 height 1080 supersample 3

12. Back to the virion

show #1 and #2 again as spheres without the cartoon and show the whole virion from the side

show #1,2 atoms
hide #1,2 cartoons
style #1,2 sphere
view

13. Add the trigger and contraction

run the script ~/phage-t4/t4_trigger.py

open ~/phage-t4/t4_trigger.py

t4_trigger.py needs t4_wave.py beside it. It fetches 9X3P, 9X3Q, 9X3R and 5IV7, fits them to the open tail and adds t4motion.

Values from 0 to 1 open the baseplate; 1 to 2 contract the sheath. Atoms matched between structures use deposited coordinates at both ends. The motion between them is modeled.

The OK card for the second script and its reply

14. Play it

switch to simple lighting so it plays smoothly, then play the baseplate trigger and the contraction over 120 frames with the t4motion command

lighting simple
t4motion play 120

Each frame moves 1.4 million atoms and the five head copies. Simple lighting avoids recalculating shadows at each frame. Playback reached about 1.5 frames per second on this PC. To record a movie, use movie record before playback and movie encode afterwards.

The contracted tail: head moved down, tube out past the star-shaped baseplate

Cost

Allow 64 MB of structure downloads and about 6 GB of free memory for the virion build. The membrane is separate.

Download or task	Download size	Time	Memory
Tail: 9MKB	34.5 MB	5 s	1.5 GB
Head: 7VS5, 5 instances	18.9 MB	5 s	+0.4 GB
Fit, `computeSS false`		11 s	3.0 GB
Fit, secondary structure recalculated (avoid)		120 s	46 GB
Motion: 9X3P, 9X3Q, 9X3R, 5IV7	11 MB	9 s	+0.4 GB
Whole build	64 MB	10 min	5.4 GB

These measurements use the PC described above. Individual steps used cached downloads and ChimeraX without a window. The whole build includes the graphics window. Memory figures are peaks; + marks an increase.

The membrane

The film includes a simulated E. coli K-12 cell envelope. Its 300 × 300 nm patch contains 32 million heavy atoms. The patch is separate from the fourteen-step build and is not available for download.

The outer membrane uses tiled all-atom snapshots, with rough LPS outside and phospholipids inside (Gutishvili 2025). OmpC trimers sit at one per 200 nm² (PDB 2J1N, oriented as in OPM). The envelope also contains one peptidoglycan layer (Boags et al. 2018) and an inner membrane (Nygaard and Nesterenko 2025). The inner membrane sits 30 nm below the outer one, as measured at T4 infection sites (Hu et al. 2015).

For your own patch, tile deposited bilayer snapshots or use CHARMM-GUI Membrane Builder (Jo et al. J Comput Chem 29:1859, 2008; Lee et al. J Chem Theory Comput 15:775, 2019). Remove molecules that overlap at tile seams and around proteins. The seams here are unrelaxed; LPS covers about 67 % of the outer leaflet.

Credits

Inspiration. Tom Goddard's ChimeraX phage T4 injection animation (RBVI, UCSF, 2022) inspired this page and its contraction-wave approach. The wave here uses separate code with T4 values.

Structures and maps (PDB and EMDB data are CC0):

9MKB, EMD-48324: Fokine A, et al. In situ structures of the portal-neck-tail complex of bacteriophage T4 inform a viral genome positioning mechanism. Nat Commun 2026. doi:10.1038/s41467-026-69106-8
7VS5: Fang Q, et al. PNAS 2022. doi:10.1073/pnas.2203272119
9X3P, 9X3Q, 9X3R: Shao Q, et al. Cryo-EM structures of phage T4 infection intermediate. J Mol Biol 438:169938 (2026). doi:10.1016/j.jmb.2026.169938
5IV7 (used only to place 9X3R): Taylor NMI, et al. Nature 2016. doi:10.1038/nature17971
2J1N: Baslé A, Rummel G, Storici P, Rosenbusch JP, Schirmer T. J Mol Biol 362:933 (2006); orientation from OPM (Lomize et al. Nucleic Acids Res 40:D370, 2012).
wwPDB consortium. Nat Struct Biol 10:980 (2003).

Dimensions and the infection sequence: Leiman PG, et al. Virol J 7:355 (2010), doi:10.1186/1743-422X-7-355; Hu B, et al. PNAS 2015, doi:10.1073/pnas.1501064112.

Membranes (CC BY 4.0, tiled): Gutishvili G. Martini 3 SI data, Zenodo 2025, doi:10.5281/zenodo.18009953; Nygaard OL, Nesterenko A. Simulation of model inner bacterial membrane (CHARMM36), Zenodo 2025, doi:10.5281/zenodo.16842791; Boags A, Samsudin F, Khalid S. Zenodo 2018, doi:10.5281/zenodo.1636577 (Boags et al. Structure 27:713, 2019).

UCSF ChimeraX. Molecular graphics and analyses performed with UCSF ChimeraX, developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, with support from National Institutes of Health. Meng EC, et al. Protein Sci 32:e4792 (2023).

Made with Pellaeon (doi:10.5281/zenodo.23395411). To cite it: Amir, Y. Pellaeon: an AI assistant for UCSF ChimeraX. Zenodo. https://doi.org/10.5281/zenodo.23395411