Protein Modeling

Protein Modeling is a two-part Division C event built around one real protein. Before the tournament, you and your partner study a crystal structure on the computer, then bend a long gray backbone by hand into the same three-dimensional shape and add the side chains and molecules that explain how it works. At the tournament you hand that model in for judging and take a written exam that tests protein chemistry, the biology of the protein's world, and your ability to explore structures in an online viewer called JUDE.

For 2027 the protein is hemagglutinin (HA) from the 2009 H1N1 pandemic influenza virus, and the part you build is the region that grabs sialic acid on the surface of human cells. That choice sets the theme for the whole event. The exam now asks how viruses infect cells and how the immune system fights back, with influenza as the worked case, and at State and Nationals it asks specifically how HA's grip on sialic acid starts an infection. Teams that score well understand one chain of cause and effect: sequence → fold → binding pocket → infection, and they can point to every link on their model.

Event format

The 2027 rules flip the event from last season's design. There is no on-the-spot building from supervisor-provided craft materials and no calculator. You build at home, from a fixed structure, to a fixed scale, and the tournament day is an exam plus a computer-exploration section.

What the rules cover

The 2027 rules manual (Protein Modeling C, Part I and Part II) divides the event into the pre-built model and the written exam. One exam topic is limited to State and National tournaments, and the model's scoring rubric changes at each level.

The material new for 2027 is the protein itself (3UBE, the 2009 pandemic HA), the influenza infection and immune response topic, the State and National sialic-acid topic, and the visualization item on ligands. The rules name a reference article for the build, "Structural Characterization of the Hemagglutinin Receptor Specificity from the 2009 H1N1 Influenza Pandemic" (Journal of Virology, DOI 10.1128/JVI.06322-11), which is free online. Read it: supervisors are likely to draw HA questions from it.

The pre-built model

The target: 3UBE, chain A, residues 153–252

3UBE is a Protein Data Bank entry for a crystal structure of the hemagglutinin from the 2009 pandemic H1N1 virus. You model only 100 residues (153 through 252, inclusive) of chain A. That stretch sits in HA1, the globular head of the protein, and it contains most of the receptor binding site. You do not model the whole trimer, the stem, or other chains in the file.

There are two sources for the coordinates. The 3D Molecular Designs Protein Modeling event page offers a modified file prepared to make modeling easier, and the original atomic coordinates are free from the RCSB Protein Data Bank. Load either into JUDE, restrict the view to the segment you are building, and switch to an alpha-carbon backbone display.

Specifications you must meet

Scale math

The alpha-carbon trace replaces each residue with one point, its α-carbon. Neighboring α-carbons are about 3.8 Å apart, because the peptide bond holds them at a fixed distance. At 2 cm per residue, every 3.8 Å of real backbone becomes 2 cm of wire. That single ratio lets you predict the size of every feature before you bend anything.

Building it, step by step

1. Map the segment. In JUDE, color by secondary structure and write down, in order from residue 153, where each helix, strand and loop starts and ends. This list is your build plan.
2. Mark the wire. Put a tick at every 2 cm, then label every tenth residue (160, 170, …). Mark the start and end of each helix and strand with tape or a marker.
3. Fold the secondary structure first. Bend helices so the center of the wire traces a circle about 2.4 cm across (a dowel about one wire-thickness narrower helps), keeping 3.6 residues per turn and a right-handed twist. Leave strands nearly straight with a slight zigzag.
4. Fold the tertiary arrangement. Rotate the model on screen until a recognizable feature (a helix, the pocket) faces you, then bend the loops so the helices and strands sit in the same places relative to each other. Check distances: if two α-carbons are 10 Å apart on screen, they should be about 5 cm apart on the model.
5. Add property colors on the α-carbons of the residues you choose to display, using the color code above.
6. Add functional features (side chains, the receptor, labels) with whatever materials you like, including 3D-printed parts.
7. Brace it. Add a stand or internal supports so it survives being lifted and rotated, and check it fits in a 50 cm cube.
8. Write the legend last, so it matches exactly what is on the model.

Functional features and the legend

Every feature you add must explain how the structure relates to function. A feature that is irrelevant, or that you cannot connect to function, earns nothing, so choose few and choose well. For the HA binding site, strong candidates are:

• The pocket floor residues whose side chains form the base of the binding site (aromatic and polar residues that hydrogen-bond to sialic acid).
• The loops and helix that frame the pocket, usually called the 130-loop, 150-loop, 190-helix and 220-loop in H3 numbering. Labeling the ones inside your segment shows judges you know where the site is.
• The receptor itself: a sialic acid, or sialic acid attached to galactose, placed in the pocket with its main contact groups indicated.
• Residues that set receptor preference, such as the aspartates at H3 positions 190 and 225 discussed below.
• Stabilizing pairs at National level: side chains whose hydrogen bonds, salt bridges or hydrophobic contacts hold the pocket's loops in place.

The Model Description Legend is a printed half sheet with a three-column table. For each feature, the first column says what it is, the second says how it is shown on the model, and the third says why it matters. A good row is specific: "Asp190 (H3 numbering) on the 190-helix / red cap on α-carbon plus printed side chain / hydrogen-bonds to the sugars of human-type α2,6 receptors, setting receptor preference." Only residues inside 153–252 have an α-carbon on your model to mark. A weak row says "important residue / red bead / helps binding."

How the model is scored

At Regionals the model is judged mainly on how accurately the alpha-carbon backbone is folded: the right helices and strands, at the right scale, packed in the right places. At State and Nationals, accuracy still carries most of the model score, but a growing slice rewards features that explain function and stability. The exam is the bigger lever at every level, so do not spend the last week before a tournament polishing the model and neglecting influenza biology.

Foundations: amino acids and side chains

Proteins are chains of amino acids. Every amino acid has a central α-carbon bonded to a hydrogen, an amino group, a carboxyl group and a side chain (the R group). Only the side chain differs among the 20 standard amino acids, and its chemistry decides where a residue sits in the fold and what it can do. That is why your model's color code is about side-chain properties, not names.

Some residues sit on the borderline. Tyrosine and tryptophan are aromatic and mostly nonpolar but carry polar groups (–OH and N–H), so different color keys place them differently; say which you chose in the legend. Glycine has only a hydrogen as its side chain and is usually grouped with the hydrophobic residues. Histidine's side chain has a pKₐ near 6, so it is only partly positive at pH 7 and becomes more positive in an acidic endosome, which matters for influenza (see the fusion section).

Charge and pH

The amino and carboxyl groups, and some side chains, gain or lose protons depending on pH. At neutral pH the carboxyl is deprotonated (–COO⁻) and the amino group is protonated (–NH₃⁺), so a free amino acid is a zwitterion: charged groups, zero net charge from the backbone. Side chains with ionizable groups add charge on top of that.

Chemistry that drives folding

The primary structure is the order of amino acids, written from the N-terminus to the C-terminus. Neighboring residues are joined by peptide bonds, which have partial double-bond character. That makes each peptide unit flat and rigid, so the backbone can only rotate at the two bonds on either side of each α-carbon. Those rotation angles, φ (phi) and ψ (psi), set the local shape of the chain, and only certain combinations avoid atoms bumping into each other. A Ramachandran plot of ψ against φ shows the allowed combinations: one region for right-handed helices, one for extended strands, and a small region on the right that mostly glycine can reach.

Secondary structure

Secondary structure is the regular local shape of the backbone, held by hydrogen bonds between backbone C=O and N–H groups.

In a beta sheet, strands lie side by side and hydrogen-bond to each other. They can be parallel (same N→C direction) or antiparallel. The receptor binding domain of HA is rich in strands and loops, with short helices, so on your model most of the wire will be strands and connecting loops rather than coils.

Tertiary and quaternary structure

Tertiary structure is the full 3D fold of one chain; quaternary structure is the assembly of several chains. HA is a trimer of three identical HA1–HA2 units, so the complete protein has quaternary structure, but your 100-residue segment shows tertiary structure within one chain.

Why proteins fold, in energy terms

Folding happens because the folded state has lower free energy than the unfolded one, ΔG = ΔH − TΔS. Unfolded chains have high conformational entropy, so folding pays an entropy cost. It is repaid by the hydrophobic effect (burying nonpolar groups releases ordered water, increasing the entropy of the solvent) and by the many hydrogen bonds, salt bridges and van der Waals contacts that form. The net stability of a typical protein is small, only tens of kJ/mol, which is why heat, extreme pH or urea can tip it over.

Anfinsen's experiment showed that a denatured enzyme (ribonuclease) refolds to its active shape when conditions are restored, so the sequence carries the information for the fold. Levinthal's paradox notes that a chain could never find its fold by random search; real chains fold quickly because they move down a funnel-shaped energy landscape toward the native state.

Denaturation unfolds a protein without breaking peptide bonds, so the primary structure survives. Whether it refolds depends on the protein. The melting temperature (Tm) is where half the molecules are unfolded and ΔG of unfolding is zero. Chaperones help large proteins fold in cells; HA itself folds and assembles into trimers in the endoplasmic reticulum with chaperone help and forms its disulfide bonds there.

Viruses and how they infect cells

A virus is a set of genes packed in a protein shell (the capsid), sometimes wrapped in a lipid envelope. It has no metabolism of its own, so it must enter a host cell and use the cell's machinery to copy itself. Every viral infection follows the same broad sequence, and exam questions often ask you to place an event in it:

1. Attachment: a viral surface protein binds a specific molecule (the receptor) on the host cell. Receptor choice sets which species and tissues the virus can infect (its tropism).
2. Entry: the virus enters by fusing its envelope with the cell membrane, or by being taken up in an endosome and escaping from it.
3. Uncoating: the genome is released from the capsid.
4. Replication and expression: viral genes are copied and translated by host ribosomes.
5. Assembly: new genomes and proteins come together.
6. Release: by budding (enveloped viruses take their envelope from the host membrane) or by bursting the cell.

Influenza virus

Structure of the virion

Influenza A is an enveloped, negative-sense single-stranded RNA virus (family Orthomyxoviridae). Its genome is split into eight segments, each wrapped with nucleoprotein (NP) and bound to its own polymerase (PB1, PB2 and PA) to form a ribonucleoprotein (vRNP). Under the envelope lies a layer of matrix protein (M1). Three proteins sit in the envelope:

The infection cycle

1. Attachment. HA on the virion binds sialic acid at the tips of glycans on airway epithelial cells.
2. Endocytosis. The cell takes the bound virus into an endosome.
3. Acidification. The endosome pumps in protons and its pH falls to about 5–6. M2 lets protons into the virion, loosening M1 from the vRNPs.
4. Fusion. Low pH triggers a large, irreversible change in HA's shape that drives the viral and endosomal membranes together (details below). The vRNPs are released into the cytoplasm.
5. Nuclear import and replication. Unlike most RNA viruses, influenza copies its genome in the nucleus. The viral polymerase makes mRNA, using 5′ caps snatched from host mRNAs to prime transcription (cap-snatching), and makes positive-sense copies that serve as templates for new genomes.
6. Translation and assembly. HA, NA and M2 travel through the ER and Golgi to the plasma membrane; new vRNPs leave the nucleus and gather under patches of membrane rich in viral proteins.
7. Budding and release. New virions bud from the cell, taking host membrane as their envelope. HA on the new virions would stick them right back onto sialic acid on the same cell, so NA cuts sialic acid to set them free.

Hemagglutinin, from precursor to fusion machine

HA is made as one chain, HA0, which folds and assembles into a trimer. A host protease must cut HA0 into HA1 and HA2, held together by a disulfide bond, before the virus can infect. Seasonal and 2009 H1N1 viruses have a single basic residue at the cleavage site, so only certain airway proteases can cut it, which keeps infection mostly in the respiratory tract. Highly pathogenic avian H5 and H7 strains have a string of basic residues there that common proteases throughout the body can cut, which is one reason they cause systemic disease.

• HA1 forms the globular head on top. It carries the receptor binding site and most of the antibody targets. Your model segment is part of HA1.
• HA2 forms most of the long stem. It holds the fusion peptide, a hydrophobic stretch at the new N-terminus created by cleavage, tucked away in the pre-fusion shape.

At endosomal pH, the HA1 heads loosen and HA2 refolds: a loop becomes part of a very long helix that fires the fusion peptide into the endosomal membrane. The helices then fold back on themselves, pulling the two membranes together until they merge. Protonation of histidines and other groups as the pH drops is thought to help trigger the change.

The 2009 H1N1 pandemic

The 2009 pandemic virus arose in pigs by reassortment, the swapping of whole genome segments between two influenza viruses that infect the same cell. Its segments came from two swine influenza lineages, which themselves carried genes of swine, avian and human origin. Its HA is an H1 from the classical swine lineage, which descends from the 1918 pandemic virus. That ancestry explains a striking feature of 2009: many older adults, exposed decades earlier to 1918-like H1 viruses, had antibodies that cross-reacted with the new HA, while children and young adults did not.

The immune response to influenza

Innate immunity

The first defenses act within hours and do not depend on prior exposure. Mucus traps virions and cilia sweep them out; mucus is also rich in sialic acid, which acts as a decoy that NA must cut through. Infected cells detect viral RNA with sensors such as RIG-I and release type I interferons, which put neighboring cells into an antiviral state. Influenza's NS1 protein counters this by blocking interferon induction. Natural killer cells kill infected cells, and macrophages and dendritic cells take up virus and debris, with dendritic cells carrying antigen to lymph nodes to start adaptive immunity. Fever, aches and fatigue largely come from the inflammatory signals of this response, not from the virus directly.

Adaptive immunity

Most neutralizing antibodies bind the HA head, around and over the receptor binding site, because it is exposed and immunodominant. On H1 heads these targets are called antigenic sites Sa, Sb, Ca1, Ca2 and Cb. The head tolerates mutation, so these sites drift. The conserved pocket floor cannot change much without losing sialic acid binding, but it is a small, shallow pocket that most antibodies cannot reach deeply. The stem is far more conserved, and broadly neutralizing stem antibodies are a major target of "universal" vaccine research.

The hemagglutination inhibition (HAI) assay measures anti-HA antibodies: HA on virions crosslinks red blood cells by binding their sialic acids (that is where the name hemagglutinin comes from), and serum antibodies that block the binding site prevent the clumping.

Prevention and treatment

Hemagglutinin and sialic acid receptors

This is the heart of the 2027 model and, at State and Nationals, a named exam topic.

Sialic acid, the receptor

Sialic acids are nine-carbon acidic sugars that cap the ends of glycan chains on cell-surface glycoproteins and glycolipids. The most common in humans is N-acetylneuraminic acid (Neu5Ac). Its key chemical features are the ones HA reads:

• a carboxylate at C1, negatively charged at physiological pH;
• an N-acetyl group at C5, whose methyl packs against a nonpolar surface;
• a glycerol side chain (C7–C9) with hydroxyls that hydrogen-bond;
• the C2 carbon, which links to the next sugar (usually galactose) through a glycosidic bond.

α2,3 versus α2,6 linkages

The bond from sialic acid's C2 to galactose can go to galactose carbon 3 (α2,3) or carbon 6 (α2,6). This one detail controls host range.

Pigs carry both linkages in their airways, which is why they can be infected by both bird and human strains and can act as a mixing vessel for reassortment, as happened in 2009.

The receptor binding site

The binding site is a shallow pocket near the top of each HA1 head. In H3 numbering, its floor is formed by a set of conserved residues, usually listed as Tyr98, Trp153, His183 and Tyr195, and its rim by four structural elements: the 130-loop, 150-loop, 190-helix and 220-loop. Sialic acid lies in the pocket with:

• its carboxylate hydrogen-bonded to backbone and side-chain groups on the 130-loop;
• its N-acetyl methyl tucked against the aromatic ring of Trp153;
• its glycerol hydroxyls hydrogen-bonded to residues near the 190-helix and the pocket floor (including Tyr98).

Each contact is weak, and a single HA–sialic acid bond has low affinity (dissociation constants in the millimolar range are commonly reported). The virus binds tightly anyway through avidity: hundreds of HA trimers on one virion bind hundreds of sialic acids at once. That is also why NA is needed, to break the bonds at the end of the cycle; flu biologists describe a balance between HA binding and NA cutting.

What sets receptor preference in H1 HAs

In human-adapted H1 viruses, including 1918 and 2009, two aspartates matter most: Asp190 and Asp225 (H3 numbering). They reach the galactose and the next sugars of an α2,6-linked glycan and make hydrogen bonds that an α2,3-linked glycan, with its different shape, does not satisfy. Avian H1 viruses usually have glutamate at 190 and glycine at 225. During 2009, a variant with glycine in place of aspartate at H3 position 225 (often called D222G in H1 numbering) was reported more often in severe cases and binds α2,3 receptors better, which may let it reach deeper lung cells. The reference article for the event compares how the 2009 HA binds human-type and avian-type receptor analogs, so read its figures with this in mind.

Working in JUDE

JUDE (Jmol User Design Environment) is the browser version of Jmol that the 2027 rules name for both the build and the exam. It is free at the 3D Molecular Designs digital modeling hub, and it is the only website you may use during the event, so practice until the common operations are muscle memory.

Element coloring follows the CPK convention: carbon gray or black, oxygen red, nitrogen blue, hydrogen white, sulfur yellow. Representation styles are often confused, so learn what each one hides: a cartoon invents ribbons that are not atoms at all, and a trace draws only α-carbons, exactly like your physical model.

Reading the chemical environment of a side chain

A common exam task gives you a residue and asks why it sits where it does. Work from the atoms outward:

1. What is the side chain made of? Carbons only means hydrophobic; N or O means hydrogen bonding; charged groups mean salt bridges.
2. What is within a few ångströms? Other side chains, backbone atoms, water, or a ligand.
3. Name the interaction and say what it stabilizes: a buried hydrophobic cluster, a surface salt bridge, a disulfide bond tying two loops together, or a contact that holds a ligand.
4. Predict a mutation's effect: putting a charged residue into a buried hydrophobic cluster destabilizes the fold; removing a cysteine breaks a disulfide; changing a pocket residue changes binding, not folding.

Mutations and sequence changes

Changing the sequence changes the protein. A missense mutation swaps one amino acid, a nonsense mutation inserts a stop codon, a silent mutation leaves the protein unchanged, and a frameshift changes everything downstream. The effect depends on where the change lands: a surface change may do nothing, a core change can destabilize the fold, and a binding-site change can alter what the protein binds.

Influenza makes this concrete. Its polymerase has no proofreading, so point mutations accumulate quickly. Changes in the antigenic sites around the HA head let the virus escape antibodies raised against last season's strain (drift). Changes inside the pocket, such as those at H3 positions 190 and 225, shift which receptor the virus prefers. Changes near the HA0 cleavage site change which proteases can activate the protein. The classic non-viral example of a single substitution with large consequences is sickle-cell disease, where glutamate at position 6 of β-globin becomes valine, creating a hydrophobic patch that makes deoxygenated hemoglobin molecules stack into fibers.

Structures, prediction and the PDB

The Protein Data Bank archives atomic coordinates of macromolecules under four-character IDs such as 3UBE. Structures come mainly from X-ray crystallography, NMR spectroscopy and cryo-electron microscopy; HA structures are usually crystallographic. A PDB file lists ATOM records with x, y and z coordinates in ångströms, HETATM records for ligands, sugars and ions, and HELIX and SHEET records naming the residues in each secondary-structure element, which is exactly the list you need for your build plan. A crystal structure is a model fitted to measured density, so flexible loops and surface glycans are often poorly ordered or missing.

AlphaFold predicts structures from sequence and reports per-residue confidence (pLDDT). A prediction is not a measurement: if a question asks how a structure was determined, the answer is an experimental method.

Strategy

• Start the model early. It must be finished before your first tournament and it travels to every level. Build it, then improve its features between tournaments as the rubric gets stricter.
• Split the preparation, not the knowledge. One partner can own the build and the JUDE operations while the other owns influenza biology and immunity, but both must be able to answer either half: the exam is 60% of the score and you sit it together.
• Put structure-specific facts on your sheets. The 3UBE segment's helix and strand boundaries, your legend's color key, the binding-site residues and loops, the α2,3 versus α2,6 contrast, the infection cycle in order, the innate and adaptive players, and the scale conversion (0.53 cm per Å).
• Rehearse JUDE under constraints. No search engine, one device, a borrowed network. Practice loading, selecting a residue range, measuring a distance and displaying a ligand without any notes.
• Protect the model in transit. It is impounded intact; a backbone that sags or a feature that falls off costs accuracy points no exam answer can win back. Build a stand and a box.
• Budget the 50 minutes. Leave a fixed block for the molecular exploration questions; they are the ones that need the device and are easiest to run out of time on.

Common pitfalls

How tournaments test this

Our archive holds 1,935 Protein Modeling questions from public tests and the practice bank holds 1,239 more; about 786 archived questions pass the site's answer-trust filter. Nearly all archived questions (1,890) come from Division C tests. Note what those numbers do and do not tell you: the archive is dominated by the 2019-20 and 2020-21 seasons (586 and 873 questions), when the event's protein was a CRISPR-associated nuclease, so a large share of archived questions are about Cas9 and base editors and are outside the 2027 scope. Use them to practice reasoning, not as a syllabus.

Question types. Of the archived questions, 54.5% are free response (1,055), 31.3% multiple choice (606), 5.7% select-all (111), 4.5% true/false (88) and 3.9% numeric (75). The 2027 rules say the exam is multiple choice and short answer, which matches the archive's balance. Expect to write sentences, and expect select-all items where more than one option is correct.

Subtopics. Tags overlap, so these add to more than 100%: Protein Structure and Folding covers 32% of archived questions, Protein Function and Binding 23%, Mutations and Sequence Biology 21%, Molecular Interactions and Stability 20%, Amino Acids and Peptide Chemistry 19%, and PDB, Jmol and Modeling 14%. The first five map directly onto the 2027 exam topics; the last maps onto the molecular visualization section.

Task profile. The most common tasks are "which of the following" (about 11%), "explain or why" (about 11%) and "calculate or compute" (about 7%). Compare-and-contrast, data interpretation and identification from a figure each stay under 2%, but figure and structure questions should rise in 2027 because the visualization section is explicitly listed.

Recurring terms. Across archive and bank prompts, the most frequent words are protein (about 600 prompts), amino (about 500), structure, acid, DNA and residues. The phrases amino acid, protein structure, alpha helix, hydrogen bonds, secondary structure, protein folding and active site recur most. Expect the 2027 crop of tests to add hemagglutinin, sialic acid, receptor binding and influenza to that list.

Most common correct answers. Keys most often name tertiary (16 questions), hydrogen bonds (14), glycine (13), proline (13), isoleucine and lysine (12 each), cysteine (11), tyrosine (11), arginine (10), quaternary and secondary (9 each). Structure-level names and residues with distinctive side chains dominate, which is a good hint about what to over-learn.

What a typical 2027 question looks like. Expect: identify a structure level or a secondary-structure element from a description or a JUDE view; pick the residue with a given property; explain how a named interaction stabilizes a fold; place a step of the infection cycle in order; explain what innate or adaptive immunity does to the virus; and, at State and Nationals, explain how HA's grip on sialic acid starts an infection and what changes receptor preference.

Test yourself

Real questions from public tournament tests in our archive, limited to the parts of the bank that stay inside the 2027 scope. Click an answer to check it; each links to the full test.

Test yourselfStructure, folding and stability

0 / 0 correct

1. 1

   How many amino acids are in each turn of a conventional alpha helix?

   Question 1 options

   A2.4B3.6C4.1D4.3

   From: USC Invitational, 2024-25, Division C

Practice more Protein Modeling questions →

Test yourselfAmino acids and side-chain chemistry

0 / 0 correct

1. 1

   TIE BREAKER: Which of the following amino acids feature positively a charged side chain at physiological pH? I. Lysine II. Aspartic Acid III. Methionine IV. Threonine

   Question 1 options

   AI onlyBII onlyCI and IIDII and IVEI,II,III, and IV

   From: University of Georgia Invitational, 2020-21, Division C

Practice more Protein Modeling questions →

Test yourselfViruses, receptors and visualization

0 / 0 correct

1. 1

   How many strands of genetic material are found inside of the influenza virus?

   Question 1 options

   A4B8C12D16

   From: Boyceville Invitational, 2017-18, Division C

Practice more Protein Modeling questions →

References and further reading

• Xu R et al., "Structural characterization of the hemagglutinin receptor specificity from the 2009 H1N1 influenza pandemic", Journal of Virology 2012, the article named in the 2027 rules: https://doi.org/10.1128/JVI.06322-11
• RCSB Protein Data Bank entry 3UBE (coordinates, ligands and structure summary): https://www.rcsb.org/structure/3UBE
• 3D Molecular Designs, Protein Modeling event page (the modified structure file and JUDE): https://learn.3dmoleculardesigns.com/science-olympiad-protein-modeling-event
• JUDE (Jmol User Design Environment), free in a browser: https://learn.3dmoleculardesigns.com/digital-modeling-hub/jude_online_jmol
• Skehel JJ and Wiley DC, "Receptor binding and membrane fusion in virus entry: the influenza hemagglutinin", Annual Review of Biochemistry 2000: https://doi.org/10.1146/annurev.biochem.69.1.531
• CDC, "How influenza viruses change: drift and shift": https://www.cdc.gov/flu/php/viruses/change.html
• NIAID, influenza overview and immune response: https://www.niaid.nih.gov/diseases-conditions/influenza
• Anfinsen CB, "Principles that govern the folding of protein chains", Science 1973: https://doi.org/10.1126/science.181.4096.223
• OpenStax, Biology 2e, chapters on biological macromolecules and on viruses and immunity: https://openstax.org/details/books/biology-2e
• Lehninger, Principles of Biochemistry, for amino-acid properties, pKₐ values and folding thermodynamics, latest edition
• Science Olympiad, Protein Modeling event page and rules: https://www.soinc.org
• SciOly Wiki, Protein Modeling: https://scioly.org/wiki/index.php/Protein_Modeling
• Practice more: the Scio.ly test archive has public Protein Modeling tests you can filter by subtopic.