We present the assessment of predictions for Template-Free Modeling in CASP10 and a report on the first ROLL experiment wherein predictions are collected year round for review at the regular CASP season. Models were first clustered so that duplicated or very similar ones were grouped together and represented by one model in the cluster. The representatives were then compared with targets using GDT_TS, QCS, and three additional superposition-independent score functions newly developed for CASP10. For each target, the top 15 representatives by each score were pooled to form the Top15Union set. All models in this set were visually inspected by four of us independently using the new plugin, EvalScore, which we developed with the UCSF Chimera group. The best models were selected for each target after extensive debate among the four examiners. Groups were ranked by the number of targets (hits) for which a group's model was selected as one of the best models. The Keasar group had most hits in both categories, with four of 19 FM and eight of 36 ROLL targets. The most successful prediction servers were QUARK from Zhang's group for FM category with three hits and Zhang-server for the ROLL category with seven hits. As observed in CASP9, many successful groups were not true "template-free" modelers but used remote templates and/or server models to obtain their winning models. The results of the first ROLL experiment were broadly similar to those of the CASP10 FM exercise.

This is a report of the assessment of the predictions made for the CASP6 protein structure prediction experiment conducted in 2004 in the New Fold (NF) category. There were nine protein domains that were judged to have new folds (NF) and 16 for which a similar structure was known but the sequence similarity was judged to be too low for them to be easily recognized (FR/A). We selected all NF targets and eight of the 16 FR/A targets judged to be at the borderline between NF and FR/A for evaluation in the NF category. A total of 165 prediction groups submitted over 7400 structural models for these targets. The quality of these models was evaluated using the GDT_TS scores of the structural similarity detection program LGA and by visual inspection of the top-scoring models. The best models submitted bore an overall similarity to the target structure for three or four of the nine NF targets and for all but one of the FR/A targets. High-scoring models for the NF targets were submitted by several different groups. When both the NF and FR/A targets were considered, Baker group dominated by submitting best models for seven of the 17 targets, but 14 other groups also managed to submit best models for one or more targets.

We present an analysis of the domain boundary prediction, a new category, in the sixth community-wide experiment on the Critical Assessment of Techniques for Protein Structure Prediction (CASP6). There were 1011 predictions submitted for 63 targets. Each prediction was compared to the set of domains defined manually by visual inspection of the experimental structure. The comparison was scored using a new domain prediction scoring scheme. As the definition of a domain is subjective, many targets were assigned alternate definitions. For such targets, each prediction was compared with all different definitions and the best score was chosen. The predictors found it difficult to accurately predict domain boundaries when the target protein contained many domains or domains made of multiple sequence segments. The CBRC-DR (P0536) and Sternberg (P0237) groups were the most successful among human experts, while Baker-Rossettadom (P0353) and Baker-Robetta-Ginzu (P0421) did well among servers.

Recombinant immunotoxins (RITs) are chimeric proteins consisting of a Fv that binds to a cancer cell and a portion of a protein toxin. One of these, Moxetumomab pasudotox, was shown to be effective in treating patients with some leukemias, where the cells are readily accessible to the RIT. However, their short half-life limits their efficacy in solid tumors, because penetration into the tumors is slow. Albumin and agents bound to albumin have a long half-life in the circulation. To increase the time tumor cells are exposed to RITs, we have produced and evaluated variants that contain either an albumin-binding domain (ABD) from or single-domain antibodies from Llama. We have inserted these ABDs into RITs targeting mesothelin, between the Fv and the furin cleavage site. We find that these proteins can be produced in large amounts, are very cytotoxic to mesothelin-expressing cancer cell lines, and have a high affinity for human or mouse serum albumin. In mice, the RIT containing an ABD from has a longer half-life and higher antitumor activity than the other two. Its half-life in the circulation of mice ranges from 113 to 194 min compared with 13 min for an RIT with no ABD. Cell uptake studies show the RIT enters the target cell bound to serum albumin. We conclude that RITs with improved half-lives and antitumor activity should be evaluated for the treatment of cancer in humans.

Immune checkpoint blockade using antibodies to cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4) benefits a limited number of cancer patients. SS1P and LMB-100 are immunotoxins that target mesothelin. We observed delayed responses to SS1P in patients with mesothelioma suggesting that anti-tumor immunity was induced. Our goal was to stimulate anti-tumor immunity by combining SS1P or LMB-100 with anti-CTLA-4. We constructed a BALB/c breast cancer cell line expressing human mesothelin (66C14-M) which was implanted in one or two locations. SS1P or LMB-100 was injected directly into established tumors and anti-CTLA-4 administered i.p. In mice with two tumors, one tumor was injected with immunotoxin and the other was not. The complete regression rate was 86% for the injected tumors and 53% for the uninjetced tumors. No complete regressions occurred when drugs were given separately. In regressing tumors, dying and dead tumor cells were intermingled with PMNs and surrounded by a collar of admixed eosinophils and mononuclear cells. Tumor regression was associated with increased numbers of tumor infiltrating CD8+ cells and blocked by administration of antibodies to CD8. Surviving mice were protected from tumor rechallenge by 66C14 cells not expressing mesothelin, indicating the development of anti-tumor immunity. The anti-tumor effect was abolished when a mutant non-cytotoxic variant was used instead of LMB-100, showing that the anti-tumor response is not mediated by recognition of a foreign bacterial protein. Our findings support developing a therapy composed of immunotoxins and checkpoint inhibitors for patients.

In CASP10, for the first time, contact-assisted structure predictions have been assessed. Sets of pairs of contacting residues from target structures were provided to predictors for a second round of prediction after the initial round in which they were given only sequences. The objective of the experiment was to measure model quality improvement resulting from the added contact information and thereby assess and help develop so-called hybrid prediction methods-methods where some experimentally determined distance constraints are used to augment de novo computational prediction methods. The results of the experiment were, overall, quite promising.

Targeting solid tumors must overcome several major obstacles, in particular, the identification of elusive tumor-specific antigens. Here, we devise a strategy to help identify tumor-specific epitopes. Glypican 2 (GPC2) is overexpressed in neuroblastoma. Using RNA sequencing (RNA-seq) analysis, we show that exon 3 and exons 7-10 of GPC2 are expressed in cancer but are minimally expressed in normal tissues. Accordingly, we discover a monoclonal antibody (CT3) that binds exons 3 and 10 and visualize the complex structure of CT3 and GPC2 by electron microscopy. The potential of this approach is exemplified by designing CT3derived chimeric antigen receptor (CAR) T cells that regress neuroblastoma in mice. Genomic sequencing of T cells recovered from mice reveals the CAR integration sites that may contribute to CAR T cell proliferation and persistence. These studies demonstrate how RNA-seq data can be exploited to help identify tumorassociated exons that can be targeted by CAR T cell therapies.

Assessment of structure predictions in CASP6 was based on single domains isolated from experimentally determined structures, which were categorized into comparative modeling, fold recognition, and new fold targets. Domain definitions were defined upon visual examination of the structures with the aid of automated domain-parsing programs. Domain categorization was determined by comparison of the target structures with those in the Protein Data Bank at the time each target expired and a variety of sequence and structure-based methods to determine potential homologous relationships.