LAP: Liability Antibody Profiler by sequence & structural mapping of natural and therapeutic antibodies.

Antibody-based therapeutics must not undergo chemical modifications that would impair their efficacy or hinder their developability. A commonly used technique to de-risk lead biotherapeutic candidates annotates chemical liability motifs on their sequence. By analyzing sequences from all major sources of data (therapeutics, patents, GenBank, literature, and next-generation sequencing outputs), we find that almost all antibodies contain an average of 3-4 such liability motifs in their paratopes, irrespective of the source dataset. This is in line with the common wisdom that liability motif annotation is over-predictive. Therefore, we have compiled three computational flags to prioritize liability motifs for removal from lead drug candidates: 1. germline, to reflect naturally occurring motifs, 2. therapeutic, reflecting chemical liability motifs found in therapeutic antibodies, and 3. surface, indicative of structural accessibility for chemical modification. We show that these flags annotate approximately 60% of liability motifs as benign, that is, the flagged liabilities have a smaller probability of undergoing degradation as benchmarked on two experimental datasets covering deamidation, isomerization, and oxidation. We combined the liability detection and flags into a tool called Liability Antibody Profiler (LAP), publicly available at lap.naturalantibody.com. We anticipate that LAP will save time and effort in de-risking therapeutic molecules.

DLAB: deep learning methods for structure-based virtual screening of antibodies.

MOTIVATION: Antibodies are one of the most important classes of pharmaceuticals, with over 80 approved molecules currently in use against a wide variety of diseases. The drug discovery process for antibody therapeutic candidates however is time- and cost-intensive and heavily reliant on in vivo and in vitro high throughput screens. Here, we introduce a framework for structure-based deep learning for antibodies (DLAB) which can virtually screen putative binding antibodies against antigen targets of interest. DLAB is built to be able to predict antibody-antigen binding for antigens with no known antibody binders. RESULTS: We demonstrate that DLAB can be used both to improve antibody-antigen docking and structure-based virtual screening of antibody drug candidates. DLAB enables improved pose ranking for antibody docking experiments as well as selection of antibody-antigen pairings for which accurate poses are generated and correctly ranked. We also show that DLAB can identify binding antibodies against specific antigens in a case study. Our results demonstrate the promise of deep learning methods for structure-based virtual screening of antibodies. AVAILABILITY AND IMPLEMENTATION: The DLAB source code and pre-trained models are available at https://github.com/oxpig/dlab-public. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

In vitro evolution of antibody affinity via insertional scanning mutagenesis of an entire antibody variable region.

We report a systematic combinatorial exploration of affinity enhancement of antibodies by insertions and deletions (InDels). Transposon-based introduction of InDels via the method TRIAD (transposition-based random insertion and deletion mutagenesis) was used to generate large libraries with random in-frame InDels across the entire single-chain variable fragment gene that were further recombined and screened by ribosome display. Knowledge of potential insertion points from TRIAD libraries formed the basis of exploration of length and sequence diversity of novel insertions by insertional-scanning mutagenesis (InScaM). An overall 256-fold affinity improvement of an anti-IL-13 antibody BAK1 as a result of InDel mutagenesis and combination with known point mutations validates this approach, and suggests that the results of this InDel mutagenesis and conventional exploration of point mutations can synergize to generate antibodies with higher affinity.

Sampling time for self-taking an oropharyngeal swab for gonorrhoea and chlamydia testing.

Self-taking oropharyngeal swabs for sexually transmitted infections such as gonorrhoea and chlamydia has become more common during the COVID-19 pandemic to minimise the risk to healthcare workers. However, there have been no standardised guidelines on sampling time for taking an oropharyngeal swab for gonorrhoea and chlamydia testing. We recruited 215 participants at the Melbourne Sexual Health Centre, Australia, between November 2021 and January 2022. We asked participants to report the time they spent on self-taking the oropharyngeal swab. The median self-taking sampling time was 8s (IQR=5-12), and the time did not differ between oropharyngeal gonorrhoea positivity (P =0.570) and oropharyngeal chlamydia positivity (P =0.457).

Clinical Presentation of Incident Syphilis Among Men Who Have Sex with Men Taking HIV Pre-Exposure Prophylaxis in Melbourne, Australia.

BACKGROUND: Current international guidelines on human immunodeficieny virus (HIV) Pre-Exposure Prophylaxis (PrEP) recommend serological screening for syphilis at routine 3-monthly PrEP appointments. The aim of our study was to describe the pattern of clinical presentation of syphilis among men who have sex with men (MSM) taking PrEP. We were interested in whether syphilis is detected through screening at scheduled3-monthly PrEP clinic appointments or whether primary or secondary syphilis presented at unscheduled interval visits. METHODS: This was a retrospective study of MSM attending the PrEP clinic at the Melbourne Sexual Health Centre between February 2016 and March 2019. Serological screening for syphilis was routinely undertaken at 3-monthly PrEP clinic appointments. Diagnoses of early syphilis were identified from PrEP clinic visits and from interim walk-in STI clinic attendances. RESULTS: There were 69 cases of early syphilis among 61 MSM taking PrEP during the study period. There were 24 (35%) primary, 16 (23%) secondary, and 29 (42%) early latent infections. The incidence of early syphilis was 8.6 per 100 person-years. A substantial proportion of primary (58%) and secondary (44%) syphilis diagnoses were made at interim STI clinic attendances, between PrEP appointments. CONCLUSIONS: Syphilis screening at routine 3-monthly PrEP visits alone fails to detect a proportion of primary and secondary syphilis infections and may be insufficient in preventing onward transmission. Education of MSM taking PrEP regarding the risk of syphilis and symptom recognition is necessary together with access to syphilis testing between PrEP visits.

Lasmiditan for the acute treatment of migraine: Subgroup analyses by prior response to triptans.

BACKGROUND: Lasmiditan demonstrated superiority to placebo in the acute treatment of migraine in adults with moderate/severe migraine disability in two similarly designed Phase 3 trials, SAMURAI and SPARTAN. Post-hoc integrated analyses evaluated the efficacy of lasmiditan in patients who reported a good or insufficient response to triptans and in those who were triptan naïve. METHODS: Subgroups of patients reporting an overall response of "good" or "poor/none" to the most recent use of a triptan at baseline (defined as good or insufficient responders, respectively) and a triptan-naïve subpopulation were derived from combined study participants randomized to receive lasmiditan 50 mg (SPARTAN only), 100 mg or 200 mg, or placebo, as the first dose. Outcomes including headache pain-freedom, most bothersome symptom-freedom, and headache pain relief 2 hours post-first dose of lasmiditan were compared with placebo. Treatment-by-subgroup analyses additionally investigated whether therapeutic benefit varied according to prior triptan response (good or insufficient). RESULTS: Regardless of triptan response, lasmiditan showed higher efficacy than placebo (most comparisons were statistically significant). Treatment-by-subgroup analyses found that the benefit over placebo of lasmiditan did not vary significantly between patients with a good response and those with an insufficient response to triptans. Lasmiditan also showed higher efficacy than placebo in triptan-naïve patients. CONCLUSIONS: Lasmiditan demonstrated comparable efficacy in patients who reported a good or insufficient response to prior triptan use. Lasmiditan also showed efficacy in those who were triptan naïve. Lasmiditan may be a useful therapeutic option for patients with migraine. TRIAL REGISTRATION: SAMURAI (NCT02439320); SPARTAN (NCT02605174).

Experimentally guided computational antibody affinity maturation with de novo docking, modelling and rational design.

Antibodies are an important class of therapeutics that have significant clinical impact for the treatment of severe diseases. Computational tools to support antibody drug discovery have been developing at an increasing rate over the last decade and typically rely upon a predetermined co-crystal structure of the antibody bound to the antigen for structural predictions. Here, we show an example of successful in silico affinity maturation of a hybridoma derived antibody, AB1, using just a homology model of the antibody fragment variable region and a protein-protein docking model of the AB1 antibody bound to the antigen, murine CCL20 (muCCL20). In silico affinity maturation, together with alanine scanning, has allowed us to fine-tune the protein-protein docking model to subsequently enable the identification of two single-point mutations that increase the affinity of AB1 for muCCL20. To our knowledge, this is one of the first examples of the use of homology modelling and protein docking for affinity maturation and represents an approach that can be widely deployed.

A global real-world assessment of the impact on health-related quality of life and work productivity of migraine in patients with insufficient versus good response to triptan medication.

BACKGROUND: Migraine is a chronic, disabling neurological disease characterized by moderate-to-severe headache pain with other symptoms, including nausea, vomiting, and photophobia. Triptans, while generally effective, are insufficiently efficacious in 30-40% of patients and poorly tolerated by or contraindicated in others. We assessed the impact of insufficient response to triptans on health-related quality of life (HRQoL) and work productivity in patients currently receiving any prescribed triptan formulation as their only acute migraine medication. METHODS: Data were from the 2017 Adelphi Migraine Disease Specific Programme, a cross-sectional survey of primary care physicians, neurologists, and headache specialists and their consulting patients with migraine in the USA, France, Germany, Italy, Spain, and UK. Triptan insufficient responders (TIRs) achieved freedom from headache pain within 2 h of acute treatment in ≤3/5 migraine attacks; triptan responders (TRs) achieved pain freedom within 2 h in ≥4/5 attacks. Multivariable general linear model examined differences between TIRs and TRs in HRQoL and work productivity. Logistic regression identified factors associated with insufficient response to triptans. RESULTS: The study included 1413 triptan-treated patients (TIRs: n = 483, 34.2%; TRs: n = 930, 65.8%). TIRs were more likely to be female (76% vs. 70% for TIRs vs TRs, respectively; p = 0.011), older (mean age 42.6 vs. 40.5 years; p = 0.003), and had more headache days/month (7.0 vs. 4.4; p < 0.001). TIRs had significantly more disability, with higher Migraine Disability Scores (MIDAS; 13.2 vs. 7.7; p < 0.001), lower Migraine-specific Quality of Life scores, indicating greater impact (Role Function Restrictive: 62.4 vs. 74.5; Role Function Preventive: 70.0 vs. 82.2; Emotional Function: 67.7 vs. 82.1; all p < 0.001), and lower EQ5D utility scores (0.84 vs. 0.91; p = 0.001). Work productivity and activity were impaired (absenteeism, 8.6% vs. 5.1% for TIRs vs. TRs; presenteeism, 34.3% vs. 21.0%; work impairment, 37.1% vs. 23.3%; overall activity impairment, 39.8% vs. 25.3%; all p < 0.05). CONCLUSION: HRQoL and work productivity were significantly impacted in TIRs versus TRs in this real-world analysis of patients with migraine acutely treated with triptans, highlighting the need for more effective treatments for patients with an insufficient triptan response. Further research is needed to establish causal relationships between insufficient response and these outcomes.

Trajectory of migraine-related disability following long-term treatment with lasmiditan: results of the GLADIATOR study.

BACKGROUND: Migraine is recognized as the second leading cause of disability globally. Lasmiditan is a novel, selective serotonin 5-HT1F receptor agonist developed for acute treatment of migraine. Here we analyzed effects of lasmiditan on migraine disability assessed with the Migraine Disability Assessment (MIDAS) scale for interim data from a long-term safety study. METHODS: Completers of two single-attack parent studies were offered participation in the 1 year GLADIATOR study, that randomized participants to treatment with lasmiditan 100 mg or 200 mg taken as needed for migraine attacks of at least moderate severity. Changes in MIDAS were modeled using a mixed model repeated measures analysis. RESULTS: The sample included 1978 patients who received ≥1 lasmiditan dose and were followed for a median of 288 days. Baseline mean MIDAS scores for the lasmiditan 100-mg and 200-mg groups were 29.4 and 28.9, respectively, indicating severe migraine-related disability. Relative to baseline, MIDAS total scores were significantly lower at 3, 6, 9, and 12 months for both dose groups. At 12 months, changes in MIDAS scores were - 12.5 and - 12.2 for lasmiditan 100 mg and 200 mg, respectively, with 49% and 53% of patients, respectively, achieving at least a 50% decrease in MIDAS total score. Statistically significant improvements were also seen for work and/or school absenteeism and presenteeism, monthly headache days, and mean headache pain intensity at all time points up to 1 year. Findings for patients who completed all visits versus those dropping out early were similar. Responses were generally similar for the lasmiditan 100 mg or 200 mg doses, between subgroups defined based on the number of baseline monthly migraine attacks (≤5 vs. >5), and also between subgroups defined by pain-free response (yes/no) during initial attacks. CONCLUSIONS: Long-term treatment with lasmiditan was associated with significant reductions in migraine-related disability, including both work or school absenteeism and presenteeism. The similarity of responses in completers and those who dropped out suggests that selective attrition does not account for the improvements. Benefits were significant at 3 months and maintained through 12 months. TRIAL REGISTRATION: clinicaltrials.govNCT02565186; first posted October 1, 2015.

Safety findings from Phase 3 lasmiditan studies for acute treatment of migraine: Results from SAMURAI and SPARTAN.

BACKGROUND: We assessed the safety profile of lasmiditan, a selective 5-HT1F receptor agonist without vasoconstrictive activity being developed as an acute therapy for migraine. METHODS: SAMURAI and SPARTAN were Phase 3 double-blind studies of patients with migraine, randomized to oral lasmiditan 50 mg (SPARTAN only), 100 mg, 200 mg, or placebo to be taken within 4 hours of onset of migraine pain. Safety data from the studies were integrated. Treatment-emergent adverse events (occurring within 48 hours of first dose) were considered in the analyses. RESULTS: The safety population comprised 1262 patients assigned placebo, and 654, 1265, and 1258 assigned lasmiditan 50 mg, 100 mg, and 200 mg, respectively. There were no deaths; serious adverse events were reported for seven patients (placebo, n = 2 [0.2%]; lasmiditan 50 mg, n = 1 [0.2%]; lasmiditan 100 mg, n = 1 [0.2%]; lasmiditan 200 mg, n = 3 [0.2%]). Patients reporting ≥ 1 treatment-emergent adverse events were: Placebo, n = 174 (13.5%); lasmiditan 50 mg, n = 166 (25.4%); lasmiditan 100 mg, n = 458 (36.2%); and lasmiditan 200 mg, n = 510 (40.6%). Treatment-emergent adverse events were generally mild or moderate in severity. The most common treatment-emergent adverse events with lasmiditan were dizziness, paresthesia, somnolence, fatigue, nausea, muscular weakness and hypoesthesia. There were no ischemic events. CONCLUSIONS: As a centrally-penetrant drug, lasmiditan use was associated with neurologic treatment-emergent adverse events; most were mild or moderate in severity and self-limiting. TRIAL REGISTRATION AT CLINICALTRIALS.GOV: SAMURAI (NCT02439320) and SPARTAN (NCT02605174).

Characterization of Dizziness After Lasmiditan Usage: Findings From the SAMURAI and SPARTAN Acute Migraine Treatment Randomized Trials.

TRIAL DESIGN: SAMURAI and SPARTAN were double-blind, placebo-controlled Phase 3 studies conducted in the United States, as well as the United Kingdom and Germany (SPARTAN only). Individuals with migraine were randomized to receive oral lasmiditan 50 mg (SPARTAN only), 100 mg, 200 mg, or placebo within 4 hours of onset of a migraine attack. The aim of this analysis was to characterize dizziness reported with lasmiditan treatment. METHODS: Data from SAMURAI and SPARTAN were pooled for the current post hoc analyses. Onset time and duration of dizziness were analyzed using descriptive statistics. Subgroup analyses based on presence/absence of dizziness were performed for the endpoints of interference with daily activity, patient global impression of change (PGIC), pain at 2 hours, and most bothersome symptom (MBS) at 2 hours based on adverse events occurring within 2 hours of taking study drug. RESULTS: Dizziness incidence was as follows: Placebo (N = 1262), 2.9% (0.1% severe); lasmiditan 50 mg (N = 654), 8.6% (0.3% severe); lasmiditan 100 mg (N = 1265), 14.9% (0.7% severe); and lasmiditan 200 mg (N = 1258), 16.8% (1.4% severe). Among participants who received lasmiditan as their first dose, risk factors for dizziness were higher lasmiditan dosage, being non-Hispanic/Latino, mild or moderate severity of migraine attack, and lower body mass index. The median time to onset of dizziness was generally 30-40 minutes, and the median duration was 1.5-2 hours. The presence of dizziness did not appear to have a negative influence on lasmiditan's effect on daily activity, PGIC, freedom from pain, or MBS. Overall, 21 participants experienced vertigo: Lasmiditan 50 mg, n = 2 (0.3%); 100 mg, n = 11 (0.9%); 200 mg, n = 7 (0.6%); and placebo, n = 1 (<0.1%). CONCLUSION: The incidence of dizziness with lasmiditan increased with dose. Dizziness was generally mild or moderate in severity and of quick onset and short duration. The presence of dizziness did not influence drug efficacy.

Monoclonal antibody targeting of fibroblast growth factor receptor 1c causes cardiac valvulopathy in rats.

Fibroblast Growth Factors (FGFs) and their receptors (FGFRs) have been proposed as potential drug targets for the treatment of obesity. The aim of this study was to assess the potential toxicity in rats of three anti-FGFR1c mAbs with differential binding activity prior to clinical development. Groups of male rats received weekly injections of either one of two FGFR1c-specific mAbs or an FGFR1c/FGFR4-specific mAb at 10 mg/kg for up to 4 weeks. All three mAbs caused significant reductions in food intake and weight loss leading to some animals being euthanized early for welfare reasons. In all three groups given these mAbs, microscopic changes were seen in the bones and heart valves. In the bones of the femoro-tibial joint, thickening of the diaphyseal cortex of long bones, due to deposition of well organized new lamellar bone, indicated that an osteogenic effect was observed. In the heart, valvulopathy described as an endocardial myxomatous change affecting the mitral, pulmonary, tricuspid and aortic valves was observed in all mAb-treated animals. The presence of FGFR1 mRNA expression in the heart valves was confirmed using in situ hybridization. Targeting the FGF-FGFR1c pathway with anti-FGFR1c mAbs leads to drug induced valvulopathy in rats. In effect, this precluded the development of these mAbs as potential anti-obesity drugs. The valvulopathy observed was similar to that described for fenfluramine and dexafenfluramine. The pathogenesis of the drug-induced valvulopathy is considered FGFR1c-mediated, based on the specificity of the mAbs and FGFR1 mRNA expression in the heart valves.

Rate of Asparagine Deamidation in a Monoclonal Antibody Correlating with Hydrogen Exchange Rate at Adjacent Downstream Residues.

Antibodies are an important class of drugs, comprising more than half of all new FDA approvals. Therapeutic antibodies must be chemically stable both in storage and in vivo, following administration to patients. Deamidation is a major degradation pathway for all natural and therapeutic proteins circulating in blood. Here, the linkage between deamidation propensity and structural dynamics is investigated by examining two antibodies with differing specificities. While both antibodies share a canonical asparagine-glycine (NG) motif in a structural loop, this is prone to deamidation in one of the antibodies but not the other. We found that the hydrogen-exchange rate at the adjacent two amides, often the autocatalytic nucleophiles in deamidation, correlated with the rate of degradation. This previously unreported observation was confirmed upon mutation to stabilize the deamidation lability via a generally applicable orthogonal engineering strategy presented here. We anticipate that the structural insight into chemical degradation in full-length monoclonal antibodies and the high-resolution hydrogen-exchange methodology used will have broad application across biochemical study and drug discovery and development.

Structural and functional characterization of a specific antidote for ticagrelor.

Ticagrelor is a direct-acting reversibly binding P2Y12 antagonist and is widely used as an antiplatelet therapy for the prevention of cardiovascular events in acute coronary syndrome patients. However, antiplatelet therapy can be associated with an increased risk of bleeding. Here, we present data on the identification and the in vitro and in vivo pharmacology of an antigen-binding fragment (Fab) antidote for ticagrelor. The Fab has a 20 pM affinity for ticagrelor, which is 100 times stronger than ticagrelor's affinity for its target, P2Y12. Despite ticagrelor's structural similarities to adenosine, the Fab is highly specific and does not bind to adenosine, adenosine triphosphate, adenosine 5'-diphosphate, or structurally related drugs. The antidote concentration-dependently neutralized the free fraction of ticagrelor and reversed its antiplatelet activity both in vitro in human platelet-rich plasma and in vivo in mice. Lastly, the antidote proved effective in normalizing ticagrelor-dependent bleeding in a mouse model of acute surgery. This specific antidote for ticagrelor may prove valuable as an agent for patients who require emergency procedures.

Evaluation of strategies to control Fab light chain dimer during mammalian expression and purification: A universal one-step process for purification of correctly assembled Fab.

Fabs are an important class of antibody fragment as both research reagents and therapeutic agents. There are a plethora of methods described for their recombinant expression and purification. However, these do not address the issue of excessive light chain production that forms light chain dimers nor do they describe a universal purification strategy. Light chain dimer impurities and the absence of a universal Fab purification strategy present persistent challenges for biotechnology applications using Fabs, particularly around the need for bespoke purification strategies. This study describes methods to address light chain dimer formation during Fab expression and identifies a novel CH 1 affinity resin as a simple and efficient one-step purification for correctly assembled Fab.

Best practices for machine learning in antibody discovery and development.

In the past 40 years, therapeutic antibody discovery and development have advanced considerably, with machine learning (ML) offering a promising way to speed up the process by reducing costs and the number of experiments required. Recent progress in ML-guided antibody design and development (D&D) has been hindered by the diversity of data sets and evaluation methods, which makes it difficult to conduct comparisons and assess utility. Establishing standards and guidelines will be crucial for the wider adoption of ML and the advancement of the field. This perspective critically reviews current practices, highlights common pitfalls and proposes method development and evaluation guidelines for various ML-based techniques in therapeutic antibody D&D. Addressing challenges across the ML process, best practices are recommended for each stage to enhance reproducibility and progress.

Machine learning designs new GCGR/GLP-1R dual agonists with enhanced biological potency.

Several peptide dual agonists of the human glucagon receptor (GCGR) and the glucagon-like peptide-1 receptor (GLP-1R) are in development for the treatment of type 2 diabetes, obesity and their associated complications. Candidates must have high potency at both receptors, but it is unclear whether the limited experimental data available can be used to train models that accurately predict the activity at both receptors of new peptide variants. Here we use peptide sequence data labelled with in vitro potency at human GCGR and GLP-1R to train several models, including a deep multi-task neural-network model using multiple loss optimization. Model-guided sequence optimization was used to design three groups of peptide variants, with distinct ranges of predicted dual activity. We found that three of the model-designed sequences are potent dual agonists with superior biological activity. With our designs we were able to achieve up to sevenfold potency improvement at both receptors simultaneously compared to the best dual-agonist in the training set.

Rapid discovery of high-affinity antibodies via massively parallel sequencing, ribosome display and affinity screening.

Developing therapeutic antibodies is laborious and costly. Here we report a method for antibody discovery that leverages the Illumina HiSeq platform to, within 3 days, screen in the order of 108 antibody-antigen interactions. The method, which we named 'deep screening', involves the clustering and sequencing of antibody libraries, the conversion of the DNA clusters into complementary RNA clusters covalently linked to the instrument's flow-cell surface on the same location, the in situ translation of the clusters into antibodies tethered via ribosome display, and their screening via fluorescently labelled antigens. By using deep screening, we discovered low-nanomolar nanobodies to a model antigen using 4 × 106 unique variants from yeast-display-enriched libraries, and high-picomolar single-chain antibody fragment leads for human interleukin-7 directly from unselected synthetic repertoires. We also leveraged deep screening of a library of 2.4 × 105 sequences of the third complementarity-determining region of the heavy chain of an anti-human epidermal growth factor receptor 2 (HER2) antibody as input for a large language model that generated new single-chain antibody fragment sequences with higher affinity for HER2 than those in the original library.

Robust prediction of relative binding energies for protein-protein complex mutations using free energy perturbation calculations.

Computational free energy-based methods have the potential to significantly improve throughput and decrease costs of protein design efforts. Such methods must reach a high level of reliability, accuracy, and automation to be effectively deployed in practical industrial settings in a way that impacts protein design projects. Here, we present a benchmark study for the calculation of relative changes in protein-protein binding affinity for single point mutations across a variety of systems from the literature, using free energy perturbation (FEP+) calculations. We describe a method for robust treatment of alternate protonation states for titratable amino acids, which yields improved correlation with and reduced error compared to experimental binding free energies. Following careful analysis of the largest outlier cases in our dataset, we assess limitations of the default FEP+ protocols and introduce an automated script which identifies probable outlier cases that may require additional scrutiny and calculates an empirical correction for a subset of charge-related outliers. Through a series of three additional case study systems, we discuss how protein FEP+ can be applied to real-world protein design projects, and suggest areas of further study.

Robust Prediction of Relative Binding Energies for Protein-Protein Complex Mutations Using Free Energy Perturbation Calculations.

Computational free energy-based methods have the potential to significantly improve throughput and decrease costs of protein design efforts. Such methods must reach a high level of reliability, accuracy, and automation to be effectively deployed in practical industrial settings in a way that impacts protein design projects. Here, we present a benchmark study for the calculation of relative changes in protein-protein binding affinity for single point mutations across a variety of systems from the literature, using free energy perturbation (FEP+) calculations. We describe a method for robust treatment of alternate protonation states for titratable amino acids, which yields improved correlation with and reduced error compared to experimental binding free energies. Following careful analysis of the largest outlier cases in our dataset, we assess limitations of the default FEP+ protocols and introduce an automated script which identifies probable outlier cases that may require additional scrutiny and calculates an empirical correction for a subset of charge-related outliers. Through a series of three additional case study systems, we discuss how Protein FEP+ can be applied to real-world protein design projects, and suggest areas of further study.