Week 4 HW: Protein Design Part I

Based on Gemini AI: 100 Dalton = 1.66054e-22 gram

Gemini AI performed the following division: 500/1.66054e-22 = 3.0110687e+24

3,011,068,700,000,000,000,000,000 amino acids in 500 grams of meat

Our cells are rebuilt from building block molecules such as amino acids. When we eat beef or fish, these proteins are broken down by our digestive system into amino acids, fatty acids, and sugars, as the beef and fish tissues are made out of them. Our body can only absorb such small molecules, and our cells make new cells based on the instructions from our DNA. Beef and fish become new human cells and tissues.

The 20 natural amino acids are not included by accident. Doig argues that specific chemical reasons are why evolution favored them, so they are included. These amino acids were selected because they enabled the formation of stable and soluble proteins (Doig 2017). The 20 natural amino acids do not have redundancy. Due to the cost of making them, redundant simple amino acids are not included evolutionarily. Occupancy of a chemical space regarding size, charge, and hydrophobicity is also not random (Philip and Freeland 2011, Ilardo et al. 2015). Kirschning’s analysis highlights amino acid biosynthesis, central metabolic networks, and separate evolutionary tracks for amino acids and cofactors (Kirschning 2022). Amino acids do not need to be optimized for redox reactions, for example. So, evolutionarily, selection is structural rather than catalytic. As reported by experimental work by Shibue et al. and Newton et al., and additionally, a theoretical work by Bywater, to achieve better folding and structurally stable proteins, 20 amino acids are needed, not fewer. Fewer amino acids would not be ideal for stability and structural diversity in protein folds (Shibue et al 2018, Newton et al. 2018, Bywater 2018). Finally, codon saturation is another factor for having the 20 natural amino acids. Although two additional amino acids, 21st (selenocysteine), and 22nd (pyrrolysine) are known to exist and be incorporated in proteins in certain living systems.

I’ve searched Google Scholar and researched Claude.

Prompts:

Why are there only 20 natural amino acids?

Can you review reference papers? Include additional insights from other papers as well. Papers were uploaded: Doig 2017 and Kirschning 2022.

For this, I’ve started by building an interactive amino acid design tool in Claude. I then ask Claude AI to design the non-natural amino acid. I focused on applying the new non-natural amino acid as a probe in a biomedical imaging application.

The designer tool enables non-natural amino acids to have multiple chemistries, including photoactivation by light that makes the amino acid bioavailable on demand and a reporter attachment for detectability.

I’ve chosen to design an imaging probe for the purpose of detecting thyroid tumors specifically.

AI will help to generate a tyrosine analog with the given considerations:

• The probe should be based on a tyrosine, as tyrosine is used along with iodine to make thyroid hormones in the thyroid. The standard amino acid backbone is suitable for uptake by the LAT1 transporter, and thyroid peroxidase (TPO) activity, which covalently attaches iodine to tyrosine, trapping it in the thyroid.

• The probe should have a one-way photo-caged/light-controlled chemistry. It is activated by light, which causes the cage to be cleaved, and tyrosine becomes available as a substrate for TPO. The circulating probe accumulates in the thyroid. Signal becomes greater than background noise, better precision detection.

• The probe should have the two-photon chromophore, as activated by NIR light, more effective in deeper tissues than a one-photon activation mechanism.

• The probe should have a radiolabelled isotope, such as ¹²³I, for traceability through isotope decay after metabolic trapping in the thyroid tissue. ¹²³I is ideal because the thyroid has a sodium-iodide symporter (NIS) that actively pumps iodide into the cells.

Fully designed result card.

I used Claude AI research to generate the interactive tool.

Prompts:

Can you build an interactive design tool to make non-natural amino acids?

Can you research why tyrosine was specifically picked for imaging the thyroid? Why not other amino acids?

Can you research the two-photon uncaging of tyrosine analogs for the detection of small tumors based on metabolic activation of tyrosine in the thyroid?

Could 123I-labelled tyrosine analogs provide new advantages over existing probes for imaging the thyroid?

D-amino acids would assume left-handed alpha-helical structures.

In D-amino acids, because of the side chain ‘R’ positioning in relation to the alpha C atom, as shown in the image below, the turn in a helical structure occurs counterclockwise, placing the side chain toward the center backbone of the helix and resulting in steric hindrance.

Although forming helices with D-amino acids is not favorable, there are still ways to achieve left-handed helices’ stability. As reported in the methods for incorporating the D-amino acids into peptides, superior performance and better therapeutic outcomes were achieved with the peptides incorporating the D-analogs (Annavarapu and Nanda 2009, Garton et al. 2018).

I’ve researched Claude.

Prompts:

Can you research D-amino acids making alpha-helices?

Explain the handedness in L and D amino acids

It is possible to discover additional helices in proteins. Other helical structures have been reported in proteins, in addition to the most common alpha helices.

As a short segment in global proteins, the 3₁₀ helix exists in nature, as demonstrated in the structure of a fungal peptide (Karle et al. 2003).

Another helix is the Pi helix, which typically forms by bulging from a long alpha helix and is associated with functional sites in proteins (Cooley et al. 2003).

A few others are polyproline II helix (PPII), polyproline I helix (PPI), and collagen triple helix, in which helical strands bundle and twist together to form a triple helix. (Hollingsworth and Karplus 2010).

A beta helix is another one that can form either left- or right-handed helices. It is a tandem protein with a repeat structure (Eisenberg 2003).

Although never observed in nature, but possible, the gamma helix was proposed as a model by Pauling et al. in 1951, and is reviewed in the historical timelines by Eisenberg (Eisenberg 2003).

I’ve researched Claude.

Prompts:

Can you research the discovery of additional helices in proteins?

What are the known helices discovered so far?

Most commonly found right-handed helices consist of left-handed amino acids and the right-handed sugars (Blackmond 2010). Biology favored this form of helix due to the structural stability provided by strong cooperativity effects of electrostatic interactions (Liu 2020).

Amino acids exist in chiral forms, L- and D-forms, where the alpha C atom has an ‘R’ side chain positioned asymmetrically. L-amino acid makes a clockwise turn in a helix, positioning the side chain ‘R’ away from the center backbone of the helix. By doing so, it avoids structural hindrance and is energetically favored. On the other hand, the positioning of the side chain in a left-handed helix would be much closer to the backbone, causing steric hindrance and not being energetically favorable.

L-amino acid became more abundant than the other chiral form, D-amino acid. There have been various hypotheses that explain this, such as in the kinetic theory, a faster production rate and slower depletion in a reaction could lead to a dominance of one form, which may have happened in the early history of the Earth.

Another is the parity violation hypothesis. In this view, although the effect imposed by the weak energy difference is very small, weak energy differences may have favored the reactions of one chirality over the other.

The most recent view considers the chirality hierarchy, which states that the chirality of helices dictates the abundance of the stereochemistry of the lower ones, amino acids, and sugars (Liu 2020).

I’ve researched Claude.

Prompts:

Can you research the reasons why most molecular helices are right-handed?

Beta-sheets formed by beta-strands are amphipathic; that is, they have two distinct behavioral characteristics: a hydrophobic core and a hydrophilic surface.

Beta-strands connect via backbone hydrogen bonds. As beta-strands lie alongside, the alpha C atoms of each strand sit across each other. Positioning of the side chains occurs in an alternating pattern, creating an amphipathic character with one side more hydrophobic and the other side more hydrophilic.

Because beta-strands lie side by side and connect, the edge strand is “exposed”. The edge strand has unsatisfied hydrogen donors and wants to connect to form hydrogen bonds. This is known as the open edge, or the “edge strand problem”. Even though a new docking strand satisfies hydrogen donors, it becomes the edge strand with unsatisfied hydrogen donors. So, due to this geometry, beta-sheets tend to form aggregates, and their open edge primarily drives protein aggregation.

Other types of interactions are also known to drive aggregation: hydrophobicity and steric zipper, which is the effect of side chains between beta-sheet proteins that create a dry interface, allowing beta-sheets to grow in parallel to the axis.

I’ve researched Claude.

Claude provided the figure.

Prompts:

Can you research the reasons beta-sheets tend to aggregate?

What is a beta-sheet protein?

Amyloid diseases, such as Alzheimer’s, are associated with the formation of fibrils and protein aggregates.

Once misfolded, the protein’s hydrophobic regions are exposed, and this triggers a self-assembly process, resulting in aggregates (amyloid fibrils) that are highly organized, rich in beta-sheet structures, and stable, achieved through an intramolecular hydrogen bonding network. In contrast, in the native state, proteins achieve stability by hydrophobic interactions via side chains.

There is no sequence dependency, as any protein can participate in hydrogen bonding. The native fold goes through misfolding and some destabilization, in which backbone hydrogen bonding can occur at this stage.

Dimerization of the monomers is the key step for locking the misfolding. As shown by Lv et al., monomers lead to cooperative formation of beta-sheet conformation and dimerization, which is thought to stabilize the misfolded state (Lv et. al. 2013).

The aggregate grows slowly until it becomes thermodynamically more stable, which occurs at a critical nucleus size (Tsemekhman et al. 2009).

Once a nucleus forms, monomers join the fibril ends rapidly, growing in parallel to the fibril axis. As in the state of an amyloid fibril, each new beta strand satisfies all hydrogen bonds. The amyloid state is thermodynamically more stable than the native state of the protein.

There are many applications reported for the use of amyloid beta-sheets. The field is expanding to include environmental remediation, biomedical, sustainable materials, and food proteins.

Amyloid fibrils provide a rigid surface and resistance to chemicals, provided by highly dense hydrogen bonds, and coded by a specific peptide sequence.

A few examples as materials:

• Bioplastic, amyloid-based material development from plant protein sources (Li et al. 2024).

• Conductive aerogels with sensing properties development. Han et al. demonstrated that in situ polymerization of amyloid fibrils as scaffolds coats the conductive polymer, polypyrrole, creating a porous, conductive network, enabling pressure sensing, strain sensing, and potential wearable electronics applications. (Han et al. 2020).

I’ve researched Claude.

Claude provided the figure.

Prompts:

Can you research why many amyloid diseases form beta-sheets?

Show me research findings about protein misfolding triggering highly organized beta-sheet formation.

What determines a misfolded protein to adopt a beta-sheet structure? What is the force behind this event?

Can you research to find examples and use cases for amyloid beta-sheets as materials?

MSSQIRQNYSTDVEAAVNSLVNLYLQASYTYLSLGFYFDRDDVALEGVSHFFRELAEEKREGYERLLKMQNQRGGRALFQDIKKPAEDEWGKTPDAMKAAMALEKKLNQALLDLHALGSARTDPHLCDFLETHFLDEEVKLIKKMGDHLTNLHRLGGPEAGLGEYLFERLTLKHD

SSQIRQNYSTDVEAAVNSLVNLYLQASYTYLSLGFYFDRDDVALEGVSHFFRELAEEKREGYERLLKMQNQRGGRALFQDIKKPAEDEWGKTPDAMKAAMALEKKLNQALLDLHALGSARTDPHLCDFLETHFLDEEVKLIKKMGDHLTNLHRLGGPEAGLGEYLFERLTL

T=0.1, sample=0, score=0.8521, seq_recovery=0.3450 GPAIRSNFSEEICAALNAQIGLERQAATTYEAMAAYFARPDVARPGVAAFFAAQAAEERAHAAALEAYLASRGCTLVETPVPAPEKAEYGDTLEAFELALAMEEEVTAAIQALIALAKANNDPETVAFFDANFVAEQAAHIAELRDYLARLRALGGPNAAEGERRFDEEVL

New Sequence:

GPAIRSNFSEEICAALNAQIGLERQAATTYEAMAAYFARPDVARPGVAAFFAAQAAEERAHAAALEAYLASRGCTLVETPVPAPEKAEYGDTLEAFELALAMEEEVTAAIQALIALAKANNDPETVAFFDANFVAEQAAHIAELRDYLARLRALGGPNAAEGERRFDEEVL