Week 4 HW: Protein Design Part I

COPT1 AlphaFold

Part A: Conceptual Questions

Answer any NINE of the following questions from Shuguang Zhang: (i.e. you can select two to skip)

1. How many molecules of amino acids do you take with a piece of 500 grams of meat? (on average an amino acid is ~100 Daltons)

6.022×1023 molecules. A 500‑gram piece of meat contains about 20% protein, so 500 g×0.20=100 g of protein. Using an average amino acid mass of roughly 100 Daltons (≈100 g/mol), this corresponds to 100 g/100 g/mol=1 mol of amino acids. Since one mole contains 6.022×1023 molecules, eating 500 grams of meat provides on the order of 6×1023 individual amino acid molecules.

2. Why do humans eat beef but do not become a cow, eat fish but do not become fish?

Humans don’t turn into cows or fish when we eat them because digestion breaks biological structures down into their simplest components: proteins are dismantled into amino acids, DNA into nucleotides, and lipids into fatty acids. Once reduced to these basic molecules, the body discards the original organism’s structure and identity and instead uses the raw materials to build human proteins and cells according to human DNA instructions. In other words, we absorb building blocks, not templates, so eating beef or fish provides nutrients but never transfers the biological “identity” of the organism we consumed.

3. Why are there only 20 natural amino acids?

There are 20 standard natural amino acids because biological evolution settled on this set that balances chemical diversity with genetic simplicity: these amino acids collectively provide a wide range of charges, sizes, polarities, and reactivities, yet they fit efficiently into the constraints of the triplet genetic code and minimize the impact of mutations by assigning similar codons to chemically similar residues. This set turned out to be chemically versatile enough to support protein folding, catalysis, and all core biochemical functions, while remaining stable, error‑tolerant, and easy for early life to encode and reproduce. Although a few additional amino acids exist, such as selenocysteine, the ~20 canonical ones represent an evolutionarily optimized core toolkit rather than a hard physical limit.

4. Can you make other non-natural amino acids? Design some new amino acids.

Yes, you can make non‑natural amino acids by designing new chemical side chains and engineering biological systems to incorporate them, which is a major focus of synthetic biology. Because amino acids are modular—each built from the same backbone but with customizable side chains—we can imagine and create new variants with functions not found in nature. Examples include bulky hydrophobic residues to stabilize protein cores, metal‑binding amino acids with tailored coordination groups, photo‑switchable residues that change shape under light, redox‑active side chains that participate in electron transfer, and rigid helix‑forming amino acids that enforce specific secondary structures. These synthetic designs expand the functional possibilities of proteins far beyond what the natural twenty amino acids can provide.

5. Where did amino acids come from before enzymes that make them, and before life started?

Amino acids existed long before enzymes or life itself because they can form through prebiotic chemistry in environments where simple molecules react under the right energy sources. Classic experiments like the Miller–Urey spark‑discharge setup demonstrated that electric energy acting on gases such as methane, ammonia, and hydrogen can generate amino acids, and similar reactions could have occurred in Earth’s early atmosphere. Additional natural sources include hydrothermal vents, where mineral‑rich, high‑temperature gradients drive organic synthesis, and meteorites such as carbonaceous chondrites, which deliver a diverse suite of amino acids formed in space. Abiotic pathways like the Strecker synthesis—where aldehydes, ammonia, and cyanide react to form amino acids—operate without biological catalysts, showing that these building blocks of life can arise spontaneously. In short, amino acids predate life, emerging from geochemical and extraterrestrial processes long before enzymes evolved to make them.

6. If you make an α-helix using D-amino acids, what handedness (right or left) would you expect?

Natural proteins are built from L‑amino acids, which form right‑handed α‑helices, and this handedness is a direct consequence of the chirality of the backbone. D‑amino acids are the mirror‑image enantiomers of L‑amino acids, so the geometry flips: every stereocenter is mirrored, and the resulting α‑helix becomes left‑handed. In other words, switching from L‑ to D‑amino acids inverts the chirality of the entire secondary structure, giving you a left‑handed α‑helix rather than the right‑handed form found in nature.

7. Can you discover additional helices in proteins?

You can discover additional helices in proteins because the α‑helix is just one of several helical solutions that polypeptide backbones can adopt. Natural proteins also contain 3₁₀ helices and π‑helices, which are less common but follow the same geometric principle of repeating backbone hydrogen bonds with different spacing and pitch. Beyond biology, synthetic systems such as foldamers and peptoids can form an even wider range of helical structures, showing that helices are a general geometric outcome of constrained backbones rather than something specific to the α‑helix or to natural amino acids.

8. Why are most molecular helices right-handed?

Most molecular helices are right‑handed because the stereochemistry of the monomers (L‑amino acids) that make up biological polymers imposes geometric and steric constraints that strongly favor a right‑handed twist. The opposite handedness would place substituents in crowded, unfavorable positions, so left‑handed helices in proteins built from L‑amino acids are rare and usually very short.

9. Why do β-sheets tend to aggregate? What is the driving force for β-sheet aggregation?

Most β‑sheets tend to aggregate because their edges expose unpaired backbone hydrogen‑bond donors and acceptors. These “sticky” edges allow β‑strands from different molecules to align and extend the hydrogen‑bonding network, creating larger intermolecular sheets. This alignment is energetically favorable because it satisfies backbone hydrogen bonds and packs side chains in a regular, low‑energy pattern. Several forces contribute to this aggregation behavior, including backbone hydrogen bonding, hydrophobic interactions between side chains, π–π stacking among aromatic residues, and the inherent planarity of the peptide backbone in β‑strands. Together, these factors make β‑sheet–rich regions prone to forming very stable, extended structures such as fibrils and aggregates, which is why β‑sheet architecture is often associated with amyloid formation.

10. Why do many amyloid diseases form β-sheets? Can you use amyloid β-sheets as materials?

Many amyloid diseases form β‑sheets because protein misfolding exposes the normally buried backbone, allowing β‑strands to align and form extended intermolecular hydrogen‑bond networks. Once a few strands stack, they create a stable template that recruits additional misfolded protein, producing long, highly ordered fibrils characteristic of amyloid aggregates. These same properties—exceptional stability, regular packing, and the ability to form long nanoscale fibers—also make amyloid‑like β‑sheets attractive as materials. Engineered amyloid fibrils can be used to build strong, tunable nanofibers with useful mechanical and chemical properties, enabling applications in biomaterials, nanotechnology, and templated assembly.

11. Design a β-sheet motif that forms a well-ordered structure.

n/a

Part B: Protein Analysis and Visualization

1. Briefly describe the protein you selected and why you selected it.

I selected Copper Transporter 1 (COPT1) from Arabidopsis thaliana (UniProt ID: Q39065, COPT1_ARATH). COPT1 is a membrane‑localized copper transporter responsible for high‑affinity copper uptake in plants. Copper is essential for processes like photosynthesis and redox chemistry but is toxic at high levels, so transporters like COPT1 are crucial for maintaining copper homeostasis. I chose this protein because it is the targeted protein for the final project on Lemna minor (duckweed), where I am interested in whether introducing or overexpressing COPT1 could enhance copper uptake for phytoremediation.

2. Identify the amino acid sequence of your protein. How long is it? What is the most frequent amino acid? You can use this Colab notebook to count the frequency of amino acids. How many protein sequence homologs are there for your protein? Hint: Use Uniprot’s BLAST tool to search for homologs. Does your protein belong to any protein family?

COPT1 is 170 amino acids long. The most frequent amino acid in the COPT1 sequence is leucine (L), which appears 18 times, making it the most abundant residue. This is consistent with COPT1 being a membrane protein, since leucine is strongly hydrophobic and commonly enriched in transmembrane helices. These homologs belong to the COPT/Ctr copper transporter family, a conserved family of high‑affinity copper transporters.

3. Identify the structure page of your protein in RCSB. When was the structure solved? Is it a good quality structure? Good quality structure is the one with good resolution. Smaller the better (Resolution: 2.70 Å). Are there any other molecules in the solved structure apart from protein? Does your protein belong to any structure classification family?

COPT1 does not have an experimentally solved structure in the PDB, but an AlphaFold model is available. Based on this model and its annotation as a membrane transporter, COPT1 belongs to the class of α‑helical membrane proteins rather than a β‑sheet–rich or globular structural family.

4. Open the structure of your protein in any 3D molecule visualization software: PyMol Tutorial Here (hint: ChatGPT is good at PyMol commands). Color the protein by secondary structure. Does it have more helices or sheets? Color the protein by residue type. What can you tell about the distribution of hydrophobic vs hydrophilic residues? Visualize the surface of the protein. Does it have any “holes” (aka binding pockets)?

COPT1 is dominated by α‑helices, especially transmembrane helices. It likely contains very few or no extended β‑sheets. So When visualized as a cartoon, the structure is dominated by α‑helices, particularly in the transmembrane region, with minimal β‑sheet content.
Coloring the protein by residue type shows that hydrophobic residues are concentrated in the transmembrane helices, forming a continuous hydrophobic band consistent with membrane insertion. Hydrophilic residues are mainly located in the cytosolic and extracellular loops and terminal regions, where they can interact with the aqueous environment. When visualized as a surface, the protein shows a predominantly hydrophobic exterior in the transmembrane region, consistent with its location in the lipid bilayer.
A central cavity is visible within the transmembrane bundle, which likely corresponds to the copper transport pathway. This cavity can be considered a binding pocket or channel through which copper ions pass.

Part C: Using ML-Based Protein Design Tools

C1. Protein Language Modeling

1. Deep Mutational Scans Use ESM2 to generate an unsupervised deep mutational scan of your protein based on language model likelihoods. Can you explain any particular pattern? (choose a residue and a mutation that stands out). (Bonus) Find sequences for which we have experimental scans, and compare the prediction of the language model to experiment.

Using ESM2 to generate an unsupervised deep mutational scan reveals clear patterns related to protein structure and biochemical constraints. For example, a mutation such as L → D (leucine to aspartic acid) in a transmembrane helix would likely have a strongly negative effect. This is because it introduces a charged residue into a hydrophobic membrane region, which is energetically unfavorable and can disrupt helix stability.

Experimental deep mutational scans of other membrane transporters show similar trends: hydrophobic residues within transmembrane helices are highly conserved, whereas loop regions tend to tolerate greater variation.

2. Latent Space Analysis Use the provided sequence dataset to embed proteins in reduced dimensionality. Analyze the different formed neighborhoods: do they approximate similar proteins? Place your protein in the resulting map and explain its position and similarity to its neighbors.

When proteins are embedded into a reduced-dimensionality latent space, each sequence is first converted by ESM2 into a high-dimensional numerical representation (embedding) that captures learned features such as structural tendencies and biochemical properties. Dimensionality reduction methods (e.g., PCA or UMAP) then project these embeddings into a lower-dimensional space while approximately preserving their relative distances. In this space, proteins with similar sequences or functions remain close together and form clusters.

Based on this, I would expect COPT1 to localize within a neighborhood composed of other plant COPT/Ctr family members. These proteins share conserved transmembrane motifs and copper-binding features, which lead to similar embeddings in the high-dimensional space. As a result, dimensionality reduction preserves their proximity, and COPT1 clusters with other high-affinity copper transporters from Arabidopsis or closely related species. This organization reflects both sequence homology and functional similarity, indicating that the latent space captures biologically meaningfulk sequence-structure-function relationships between proteins.

C2. Protein Folding
Folding a protein. Fold your protein with ESMFold. Do the predicted coordinates match your original structure? Try changing the sequence, first try some mutations, then large segments. Is your protein structure resilient to mutations?

The ESMFold prediction would likely resemble the AlphaFold model: a bundle of α‑helices forming a membrane‑spanning structure. In terms of mutations, the predicted structure is expected to be relatively resilient to small changes, particularly in loop regions, which are more flexible and tolerate variation without significantly altering the overall fold. However, mutations within transmembrane helices—especially those that replace hydrophobic residues with polar or charged ones—would likely destabilize the structure, as they disrupt favorable interactions within the membrane environment.

Larger sequence changes, such as altering entire segments, would likely have a much more significant impact on folding. Membrane proteins depend on precise helix packing and specific residue interactions, so disrupting these features would likely lead to major structural changes or misfolding.

C3. Protein Generation
Inverse-Folding a protein: Let’s now use the backbone of your chosen PDB to propose sequence candidates via ProteinMPNN. Analyze the predicted sequence probabilities and compare the predicted sequence vs the original one. Input this sequence into ESMFold and compare the predicted structure to your original.

ProteinMPNN would propose sequences that maintain hydrophobic residues in the transmembrane helices and preserve key motifs required for copper transport. The predicted sequence would likely be similar to the original, especially in the membrane‑spanning regions where structural constraints are strongest. The predicted sequence would probably match the original at many positions, particularly in the helices. Differences would appear mostly in loop regions, where the structure is less constrained. Folding the ProteinMPNN‑generated sequence with ESMFold would likely produce a structure similar to the original helical bundle.

Part D. Group Brainstorm on Bacteriophage Engineering

1. Find a group of ~3–4 students
2. Read through the Phage Reading material listed under “Reading & Resources” below.
3. Review the Bacteriophage Final Project Goals for engineering the L Protein:

• Increased stability (easiest)
• Higher titers (medium)
• Higher toxicity of lysis protein (hard)

4. Brainstorm Session

• Choose one or two main goals from the list that you think you can address computationally (e.g., “We’ll try to stabilize the lysis protein,” or “We’ll attempt to disrupt its interaction with E. coli DnaJ.”).
• Write a 1-page proposal (bullet points or short paragraphs) describing:
  • Which tools/approaches from recitation you propose using (e.g., “Use Protein Language Models to do in silico mutagenesis, then AlphaFold-Multimer to check complexes.”).
  • Why do you think those tools might help solve your chosen sub-problem?
  • Name one or two potential pitfalls (e.g., “We lack enough training data on phage–bacteria interactions.”).
  • Include a schematic of your pipeline.
• This resource may be useful: HTGAA Protein Engineering Tools

5. Each individually put your plan on your HTGAA website

• Include your group’s short plan for engineering a bacteriophage