Yao-ming Huang

Quantitative in vivo solubility and reconstitution of truncated circular permutants of green fluorescent protein

Several versions of split green fluorescent protein (GFP) fold and reconstitute fluorescence, as do many circular permutants, but little is known about the dependence of reconstitution on circular permutation. Explored here is the capacity of GFP to fold and reconstitute fluorescence from various truncated circular permutants, herein called “leave‐one‐outs” using a quantitative in vivo solubility assay and in vivo reconstitution of fluorescence. Twelve leave‐one‐out permutants are discussed, one for each of the 12 secondary structure elements. The results expand the outlook for the use of permuted split GFPs as specific and self‐reporting gene encoded affinity reagents.

Green‐lighting green fluorescent protein: Faster and more efficient folding by eliminating a cis–trans peptide isomerization event

Wild‐type green fluorescent protein (GFP) folds on a time scale of minutes. The slow step in folding is a cis–trans peptide bond isomerization. The only conserved cis‐peptide bond in the native GFP structure, at P89, was remodeled by the insertion of two residues, followed by iterative energy minimization and side chain design. The engineered GFP was synthesized and found to fold faster and more efficiently than its template protein, recovering 50% more of its fluorescence upon refolding. The slow phase of folding is faster and smaller in amplitude, and hysteresis in refolding has been eliminated. The elimination of a previously reported kinetically trapped state in refolding suggests that X‐P89 is trans in the trapped state. A 2.55 Å resolution crystal structure revealed that the new variant contains only trans‐peptide bonds, as designed. This is the first instance of a computationally remodeled fluorescent protein that folds faster and more efficiently than wild type.

Improving computational efficiency and tractability of protein design using a piecemeal approach. A strategy for parallel and distributed protein design

MOTIVATION Accuracy in protein design requires a fine-grained rotamer search, multiple backbone conformations, and a detailed energy function, creating a burden in runtime and memory requirements. A design task may be split into manageable pieces in both three-dimensional space and in the rotamer search space to produce small, fast jobs that are easily distributed. However, these jobs must overlap, presenting a problem in resolving conflicting solutions in the overlap regions. RESULTS Piecemeal design, in which the design space is split into overlapping regions and rotamer search spaces, accelerates the design process whether jobs are run in series or in parallel. Large jobs that cannot fit in memory were made possible by splitting. Accepting the consensus amino acid selection in conflict regions led to non-optimal choices. Instead, conflicts were resolved using a second pass, in which the split regions were re-combined and designed as one, producing results that were closer to optimal with a minimal increase in runtime over the consensus strategy. Splitting the search space at the rotamer level instead of at the amino acid level further improved the efficiency by reducing the search space in the second pass. AVAILABILITY AND IMPLEMENTATION Programs for splitting protein design expressions are available at www.bioinfo.rpi.edu/tools/piecemeal.html CONTACT: bystrc@rpi.edu Supplementary information: Supplementary data are available at Bioinformatics online.

Expanded Explorations into the Optimization of an Energy Function for Protein Design

Nature possesses a secret formula for the energy as a function of the structure of a protein. In protein design, approximations are made to both the structural representation of the molecule and to the form of the energy equation, such that the existence of a general energy function for proteins is by no means guaranteed. Here, we present new insights toward the application of machine learning to the problem of finding a general energy function for protein design. Machine learning requires the definition of an objective function, which carries with it the implied definition of success in protein design. We explored four functions, consisting of two functional forms, each with two criteria for success. Optimization was carried out by a Monte Carlo search through the space of all variable parameters. Cross-validation of the optimized energy function against a test set gave significantly different results depending on the choice of objective function, pointing to relative correctness of the built-in assumptions. Novel energy cross terms correct for the observed nonadditivity of energy terms and an imbalance in the distribution of predicted amino acids. This paper expands on the work presented at the 2012 ACM-BCB.

Toward Computationally Designed Self-Reporting Biosensors Using Leave-One-Out Green Fluorescent Protein

Leave-one-out green fluorescent protein (LOOn-GFP) is a circularly permuted and truncated GFP lacking the nth β-strand element. LOO7-GFP derived from the wild-type sequence (LOO7-WT) folds and reconstitutes fluorescence upon addition of β-strand 7 (S7) as an exogenous peptide. Computational protein design may be used to modify the sequence of LOO7-GFP to fit a different peptide sequence, while retaining the reconstitution activity. Here we present a computationally designed leave-one-out GFP in which wild-type strand 7 has been replaced by a 12-residue peptide (HA) from the H5 antigenic region of the Thailand strain of H5N1 influenza virus hemagglutinin. The DEEdesign software was used to generate a sequence library with mutations at 13 positions around the peptide, coding for approximately 3 × 10(5) sequence combinations. The library was coexpressed with the HA peptide in E. coli and colonies were screened for in vivo fluorescence. Glowing colonies were sequenced, and one (LOO7-HA4) with 7 mutations was purified and characterized. LOO7-HA4 folds, fluoresces in vivo and in vitro, and binds HA. However, binding results in a decrease in fluorescence instead of the expected increase, caused by the peptide-induced dissociation of a novel, glowing oligomeric complex instead of the reconstitution of the native structure. Efforts to improve binding and recover reconstitution using in vitro evolution produced colonies that glowed brighter and matured faster. Two of these were characterized. One lost all affinity for the HA peptide but glowed more brightly in the unbound oligomeric state. The other increased in affinity to the HA peptide but still did not reconstitute the fully folded state. Despite failing to fold completely, peptide binding by computational design was observed and was improved by directed evolution. The ratio of HA to S7 binding increased from 0.0 for the wild-type sequence (no binding) to 0.01 after computational design (weak binding) and to 0.48 (comparable binding) after in vitro evolution. The novel oligomeric state is composed of an open barrel.

Exploring objective functions and cross-terms in the optimization of an energy function for protein design

Nature possesses a secret formula for the energy as a function of the structure of a protein. In protein design, approximations are made to both the structural representation of the molecule and to the form of the energy equation, such that the existence of a general energy function for proteins is by no means guaranteed. Here we present new insights towards the application of machine learning to the problem of finding a general energy function for protein design. Machine learning requires the definition of an objective function, which carries with it the implied definition of success in protein design. We explored four functions, consisting of two functional forms, each with two criteria for success. Optimization was carried out by a Monte Carlo search through the space of all variable parameters. Cross-validation of the optimized energy function against a test set gave significantly different results depending on the choice of objective function, pointing to relative correctness of the built-in assumptions. Novel energy cross-terms correct for the observed non-additivity of energy terms and an imbalance in the distribution of predicted amino acids.