Evaluation of the Feasibility of Screening Tau Radiotracers Using an Amyloid Biomathematical Screening Methodology

The purpose of this study is to evaluate the feasibility of extending a previously developed amyloid biomathematical screening methodology to support the screening of tau radiotracers during compound development. 22 tau-related PET radiotracers were investigated. For each radiotracer, in silico MLogP, V x, and in vitro K D were input into the model to predict the in vivo K 1, k 2, and BPND under healthy control (HC), mild cognitive impaired (MCI), and Alzheimer's disease (AD) conditions. These kinetic parameters were used to simulate the time activity curves (TACs) in the target regions of HC, MCI, and AD and a reference region. Standardized uptake value ratios (SUVR) were determined from the integrated area under the TACs of the target region over the reference region within a default time window of 90–110 min. The predicted K 1, k 2, and BPND values were compared with the clinically observed values. The TACs and SUVR distributions were also simulated with population variations and noise. Finally, the clinical usefulness index (CUI) ranking was compared with clinical comparison results. The TACs and SUVR distributions differed for tau radiotracers with lower tau selectivity. The CUI values ranged from 0.0 to 16.2, with 6 out of 9 clinically applied tau radiotracers having CUI values higher than the recommend CUI value of 3.0. The differences between the clinically observed TACs and SUVR results showed that the evaluation of the clinical usefulness of tau radiotracer based on single target binding could not fully reflect in vivo tau binding. The screening methodology requires further study to improve the accuracy of screening tau radiotracers. However, the higher CUI rankings of clinically applied tau radiotracers with higher signal-to-noise ratio supported the use of the screening methodology in radiotracer development by allowing comparison of candidate radiotracers with clinically applied radiotracers based on SUVR, with respect to binding to a single target.


Introduction
Alzheimer's disease (AD) is a progressive neurodegenerative disorder defined by histopathological features such as senile plaques and neurofibrillary tangles (NFT), and clinical symptoms such as memory loss and reduced executive functions [1]. e yearly number of AD cases is increasing worldwide, leading to an increased cost of care for dementia patients. Positron emission tomography (PET) using amyloid and tau radiotracers can measure the amyloid and tau loads, in terms of standardized uptake values ratio (SUVR), and their distributions in a subject's brain from static PET images. Since abnormal accumulation of amyloid and tau in the brain occurs before clinical symptoms appear, the imaging of these precursors can support differential diagnosis and early intervention to increase the success rate of treating AD or slow down the rate of dementia. As such, the 2018 National Institute on Aging-Alzheimer's Association (NIA-AA) research framework includes not only symptomatic stages of AD, but also biomarker classification involving amyloid, tau, and neurodegeneration AT(N) biomarkers [2]. e new framework will be able to identify subjects at risk for AD for suitable and early treatment, in particular, preclinical AD subjects (classified as A+T−(N−) or A+T+(N+)), who are not cognitively impaired but have abnormal amyloid and tau protein deposits [2].
Despite active efforts since 2000 to develop amyloid and tau-targeting PET radiotracers to assist in the diagnosis of AD and to support AD drug development, there are few radiotracers that have made it into clinical studies and displayed good clinical efficacy. In conventional radiotracer and drug development, poor bench-to-bedside translation often results due to the differences between in vitro and in vivo conditions. Similarly, animal models, especially rodents, are often poor predictors of human physiology and treatment response and have been reported to be incorrect in approximately one out of three cases [3]. Although larger animals (e.g., pigs and primates) show closer physiology to that of human, they are still in-prefect human models and are costly for high-throughput screening compared to rodents. ese issues lead to high attrition rates in drug and radiotracer development. Biomathematical simulation can complement high-throughput screening by allowing simultaneous and rapid evaluation of many candidate radiotracers [4][5][6].
Compared to amyloid radiotracers, the development of a successful tau radiotracer encounters additional challenges due to the tau phenotypes. Tau proteins have six isoforms, which differ in the number of exons (0, 1, 2) on the acidic region and the number of repeats (3 repeats (3R) or 4R) in the repeat-domain regions [7]. e different isoforms undergo several posttranslational modifications, leading to various ultrastructural conformations, which will affect the binding of tau radiotracers. In addition, they also need to discriminate between the paired helical filament (PHF) tau from other β-sheet structured aggregates such as amyloidbeta (Aβ) and α-synuclein. Although the tau protein is larger than the Aβ protein, the tau binding sites are present in smaller concentrations compared to the Aβ binding sites by 5-20 folds; hence, the selectivity of tau over other β-sheet structured aggregates needs to be high to ensure accurate quantification. Moreover, as tau proteins exist intracellularly, tau radiotracers not only need to cross the blood-brain barrier (BBB), they also need to be able to cross the cell membrane [8].
Existing clinically applied tau radiotracers showed some limitations. [ 11 C]PBB3 has high binding selectivity to tau over Aβ, but it is difficult to synthesize as it will undergo photoisomerization [9]. Moreover, it is rapidly metabolized in the plasma, and its polar metabolite is shown to cross the blood-brain barrier and enter into the brain [10]. e short half-life of carbon-11 has also prompted the development of fluorinated PBB3 compounds ([ 18 F]AM-PBB3 and [ 18 F]PM-PBB3) and other tau radiotracers so that they can be used in hospitals without dedicated cyclotron facilities. [ 18 F]T808 (also known as [ 18 F]AV-680) exhibits defluorination, which will affect the quantitative analysis of PET images especially for regions near the skull [11]. Some THK compounds (Tohoku University, Japan) showed differences in the uptake due to the enantiomeric properties of the compounds [12]. A serious confounding factor facing the development of tau radiotracers is off-target brain binding, which might affect the quantitative analysis of the PET images as observed in [  was reported to have reduced off-target binding but further evaluation was still required [16].
We have previously developed an amyloid biomathematical screening methodology to support the screening of candidate amyloid radiotracers during compound development [4,5]. e screening methodology predicts the standardized uptake values ratios (SUVRs) of different subject conditions of a radiotracer and then compares the clinical usefulness of multiple radiotracers simultaneously in discriminating the subject conditions using a clinical usefulness index (CUI). e CUI was developed to objectively evaluate the clinical usefulness of a radiotracer, based on its binding capability to a single target of interest, in terms of SUVR. e SUVR is a semiquantitative parameter that generalizes the complicated behaviors of tau radiotracers. SUVR is also generally preferred for diagnosis of patients in amyloid and tau imaging; hence, the clinical data are more readily available for comparison.
us, we chose SUVR over other kinetic parameters such as nondisplaceable binding potential (BP ND , unitless).
In this study, we evaluate the feasibility of extending the amyloid-validated screening methodology to support the development of tau PET radiotracers, where more challenges like off-target binding exist. is is the first in silico method investigated, which uses the physicochemical and pharmacological properties of the compounds to support tau PET radiotracers developments. 22 PET radiotracers reported to bind to tau proteins were investigated, including 9 clinically applied and tau-focused radiotracers, namely, [ 18

Materials and Methods
An overview of the amyloid biomathematical methodology is described briefly, followed by the screening of tau PET radiotracers using the biomathematical methodology. e details of the methodology are found in somewhere [4,5].

Biomathematical Screening Methodology.
e screening methodology was based on a simplified 1-tissuecompartment model (1TCM), with the assumption that the radiotracers cross the blood-brain barrier (BBB) by passive diffusion. It consists of four main parts ( Figure 1).

Generation of Physicochemical and Pharmacological
Parameters. A total of three inputs were required for each radiotracer: in silico molecular volume and lipophilicity as represented by McGowan Volume (V x , cm 3 /mol/100), Moriguchi LogP (MLogP, unitless), and an in vitro dissociation constant (K D , nM) ( Table 1). V x and MLogP were generated based on the chemical structure of the radiotracer using commercial software, dproperties (Talete, Italy). K D values were extracted from the literature, measured via binding assays, using synthetic tau or human brain homogenates. MLogP was used to derive the free fractions of the radiotracer in tissues (f ND , unitless) and in plasma (f P , unitless) from the following relationships [4]: (1) e list of 22 tau radiotracers and their respective inputs are shown in Table 1. e K D values that were utilized for simulations are given in bold for human brain homogenates, and italicized for synthetic tau, if available for comparison.

Derivation of 1TCM Kinetic Parameters.
e influx rate constant (K 1 , mL/cm 3 /min) was derived using the modified Renkin and Crone equation, using compound-specific permeability (P, cm/min), with fixed values of capillary surface area (S � 150 cm 2 /cm 3 of brain) and perfusion (f � 0.6 mL/cm 3 /min) as follows [4,6]: e compound-specific permeability was derived from the simplified Lanevskij's permeability model, with MLogP and V x as inputs [4,6]: e efflux rate constant (k 2 , min −1 ) can be derived using K 1 , f P , and f ND at equilibrium: e in vivo nondisplaceable binding potential (BP ND , unitless) was determined using Mintun's equation with B avail , f ND, and K D : e available tau-binding sites (B avail , nM) were measured using enzyme-linked immunosorbent assay (ELISA). e total amount of tau fibrils (B avail , nM) in the frontal lobes, parietal lobes, and hippocampus in HC and AD were 1.5 and 16.0 nM, respectively [29], assuming a tau molecular weight of 78,928 Da (https://www.phosphosite.org).

Simulations of Population Time Activity Curves (TACs) and SUVRs.
e predicted K 1 , k 2 , and BP ND were used to Influx rate constant Efflux rate constant Binding potential In silico An input function with similar kinetics to that observed in tau imaging with a fast uptake and washout is required to reflect tau kinetics. For our simulations, a fixed arterial input function was applied with fast kinetics that was derived by averaging the metabolite-corrected arterial plasma input functions of 6 HC subjects injected with [ 11 C]BF227 [30]. e same K 1 and k 2 scaling factors of 1.23 and 1.15, respectively, were introduced to account for the differences between the predicted and in vivo values [5]. e scaling factor of BP ND was modified from 0.39 to 1.0 because there were few reported values to determine the appropriate scaling factor. Monte Carlo simulations were applied to generate 1000 TACs in both target and reference regions with 3% noise, to reflect the noise in PETdata, and the population variation, by varying K 1 and k 2 by 10% and 20%, respectively [5,6]. e variations in the tau fibrils in HC and AD were determined as 10% and 35%, respectively, using the ratio of the summed standard deviation to the mean value [29]. e amount of soluble tau in HC, MCI, and AD was reported, but since they did not correlate well with the amount of phosphorylated tau, these values could not be used [31]. In our simulations, the total amount of tau fibrils in MCI was assumed to be the mean of that in HC and AD, with the same amount of variation of 35%, as used for the amyloid simulations [5].
1000 noisy TACs in both target and reference regions were generated by computer simulations with noise. In our simulation, the target region refers to a brain region with varying concentrations of phosphorylated tau depending on subject conditions (e.g., temporal lobe) and a reference is a brain region devoid of phosphorylated tau (e.g., cerebellum). 1000 SUVRs of each subject condition of HC, MCI, and AD were determined from the ratio of the areas under the TACs of the target regions in HC, MCI, and AD and that of the reference region within a chosen time window. For our simulations, a default time window of 90-110 min was selected as the predicted TACs of HC, MCI, and AD appeared to reach a quasi-steady-state in this time window for almost all 9 clinically applied tau radiotracers (Supplementary 2). To evaluate the efficacy of fixed time windows, SUVRs were also determined using the literature-reported time windows for the 9 clinically applied radiotracers. Table 1: In silico MLogP and V x and in vitro K D of 22 tau-related PET radiotracers. K D values employed for simulations are given in bold (measured using brain homogenates) and italicized (measured using synthetic tau).

Radiotracers
MLogP  [28] Units: MLogP (unitless), V x (cm 3 /mol/100), K D (nM). $ Averaged K D values (2.2, 3.1) for tau in AD brain homogenates of temporal and hippocampus. β Averaged K D values 0.14, 0.30, 0.25, 0.24, and 0.38 for tau in AD brain homogenates of frontal and entorhinal cortex of 5 AD. α K D values are measured using synthetic tau (K18Δ280K) & K i values measured using AD brain homogenates with THK5105 as competitor δ K i values measured using AD brain homogenates with T808 as competitor. # K D values measured using AD brain via autoradiography.
e simulated TACs and the predicted SUVR were compared to the clinical data of 9 clinically applied tau radiotracers.
e predicted K 1 , k 2 and BP ND values were compared with the clinically observed values where applicable. Finally, the list of 22 tau radiotracers ( Table 1) was evaluated using CUI. We previously developed a MATLABbased program, RSwCUI, (Ver. 2014b, e MathWorks, US) [5], to support the screening of amyloid radiotracers based on the proposed amyloid biomathematical screening methodology. e program was used for the evaluation of tau radiotracers in this study. Figure 2 shows the simulated TACs for the target regions of HC, MCI, and AD and reference regions of 9 clinically applied tau radiotracers. In general, the clinically observed TACs of THK compounds of the reference region had higher peaks and faster washout in the cerebellum than the target regions [15,[32][33][34][35], while the peaks of the simulated TACs of the reference region were always lower than that of the target regions (Figures 2(a)-2(e)). e simulated TACs of [ 11 C] PBB3 ( Figure 2(f )) were close to that observed clinically in AD in the nonbinding and low-, middle-, and high-binding regions [10].

Results
e simulated TACs of [ 18 F]flortaucipir (Figure 2(g)) had slightly sharper peaks and faster washout compared to the clinically observed TACs for both HC and AD [36]. Unlike the THK compounds, the peaks of the clinically observed TACs of the target regions of [ 18 F]flortaucipir were higher than that of the reference region, which was also observed in the simulated TACs [36]. e predicted TACs of [ 18 F]T808 for both the reference and the target regions of HC, MCI, and AD conditions completely overlapped with each other (Figure 2(h)). e clinically observed TACs of [ 18 F]T808 appeared close to that of [ 18 F]flortaucipir, but with smaller differences between the subject conditions. However, the simulated TACs showed complete overlapped between the HC and AD conditions with a slower uptake and washout [37]. e predicted TACs of both target and reference regions of [ 18 F]MK6240 showed similar fast uptake but slower washout than clinically observed TACs [16]. Table 2 compares the predicted and clinically-reported values of K 1 , k 2 , and BP ND of five clinically applied tau radiotracers with reported kinetic parameters. For [ 18 F] flortaucipir, the predicted K 1 and k 2 values of 0.256 and 0.199, respectively, were relatively close to the reported averaged cerebellar K 1 and k 2 values of 0.26 and 0.17, respectively [36]. e predicted k 2 value of [ 18 F]THK5351 was 0.140, which was higher than the clinically observed value of 0.115, with a difference of 21.7% [38]. However, unlike [ 18 F] flortaucipir where both K 1 and k 2 values were determined using the two-tissue-compartment model with a variable fraction [36], the reported k 2 value of [ 18 F]THK5351 was an apparent rate constant from reference region to plasma, which was determined using the simplified reference tissue model (SRTM) [38].  [16]. Table 3 shows the predicted SUVR values obtained using the default time window and literature-reported time window of 90-110 min, and the clinically observed SUVR for 10, 10, and 9 clinically applied tau radiotracers. e differences in the SUVRs predicted using both time windows were very small for both HC and AD. e predicted SUVR for HC was always greater than 1.0, but the clinically observed SUVR values were less than 1.0 for some radiotracers. In general, the clinically observed SUVR for HC and AD were greater than the predicted SUVR determined using the literature-reported time window, except for [ 11 C]PBB3 and [ 18 F]MK6240, where the predicted SUVR for HC and AD were greater. e correlations between the predicted and highest clinically observed SUVR for AD were similar with coefficients of determination, R 2 of 0.90 and 0.89, respectively, using the literature-reported time window and the default time window (Figure 3). However, the good correlation was driven by [ 18 F]MK6240, which had the highest predicted and clinically observed SUVR. Poor correlation was observed after removing [ 18 F]THK5351 and [ 18 F]MK6240. e small difference between the predicted SUVR using the default and clinical-reported time window, and the value of R 2 , showed that the default time window of 90-110 min was suitable for predicting the SUVR of the tau radiotracers (Figure 3). e simulated SUVR distribution of [ 18 F]THK523 across HC, MCI, and AD conditions substantially overlapped each other (Figure 4(a)). However, the clinically observed SUVR distribution of [ 18 F]THK523 differed across different regions of interest, with HC− (PIB-negative) having the smallest spread and smallest values, HC+ (PIB-positive) having a relatively large spread and values ranging between that of HC− and AD, and AD subjects having the largest values and a nearly similar spread as HC+ [30]. For [ 11 (Figures 4(b)-4(d)). is supported the use of 35%, 35% and 10% variations in B avail for population simulations. Figure 5 shows the CUI distribution of 22 tau-related radiotracers. Among the clinically applied tau radiotracers, [

Discussion
In this paper, we evaluated the feasibility of extending a previously developed amyloid biomathematical screening   (Figure 4). Both the predicted and clinically observed SUVR values were less than 1.0 in HC for some radioligands, especially those with a lower selectivity for tau (e.g., [ 18 F]THK523). e clinically observed SUVR of AD is much higher than that of HC. However, there is little difference in the predicted SUVR for is shows that the predictions were less accurate for tau compounds with a lower selectivity for the target. Poor predictions might be due to binding to other β-sheet structured proteins or off-target sites shown in the clinical data, whereas the predicted values showed the binding of the radiotracers to only the target site. Nonspecific binding in white matter may also lead to spill-over into the surrounding cortical regions, leading to higher clinically observed SUVRs. e issue of non-specific binding   (Table 2). e predictions for BP ND were generally poor for the three clinically-reported tau radiotracers (Table 2). is may be due to the use of a simplified 1TCM for prediction, even though 2TCM was reported to be more suitable for many clinically applied tau radiotracers. e simplified 1TCM was selected even though 2TCM is more accurate for modeling tau kinetics as the prediction of a larger number of microparameters may be difficult to estimate reliably. Moreover, the 1TCM worked reasonably   well in predicting the kinetics of the amyloid radiotracers, even though 2TCM was reported to be more suitable [5]. Other possible reasons for the poorer BP ND predictions included differences in binding to the plasma proteins due to the enantiomeric properties of the radiotracers [42], metabolites crossing the BBB for [ 11 C]PBB3 [10], binding of tau radiotracers to other similar β-sheet structures (Aβ and α-synuclein), or off-target binding in target regions of interest [13][14][15]. e predicted 1TCM parameters and SUVR, as well as the simulated TACs and SUVR distribution, were compared to clinically observed data where applicable. However, we were limited by the small number of reported kinetic parameters and SUVR values to fully assess the amyloid biomathematical model for screening tau radiotracers. e predicted and highest clinicallyobserved SUVR data for AD correlated well using fixed time window of 90-110 min and the literature-reported time window with R 2 values of 0.88 and 0.89 respectively, for 9 clinically applied tau radiotracers (Figure 3). However, the results were driven mostly by [ 18 F]MK6240. Some of the clinically applied tau radiotracers ([ 18 F]THK523, [ 18 F]THK5351 and [ 18 F]flortaucipir) did not have high selectivity for tau, which may have contributed to smaller predicted values as the predicted values were based on binding to a single target site but the off-target binding or specific binding to other β-sheet structures (e.g., amyloid) may yield higher clinical SUVR values. e predicted TACs of [ 18 F]T808 exhibited a much slower clearance compared to the clinically observed kinetics, which resulted in a large difference between the predicted and clinically observed SUVR. is may be due to the poor predictive ability of in silico parameters for [ 18 F] T808, which has a unique chemical structure. C]Astemizole yielded small CUI values using the K D values measured using human brain homogenates, which differed greatly from that measured using synthetic tau. K D or K i values measured using AD brain homogenates were very different from those measured using heparin-induced tau polymer (HITP) ( Table 1).

Comparison of Tau
is is because HITP is composed of only 3R and/or 4R, and hence may not undergo the same phosphorylation process as human tau [19,43]. On the other hand, the K D or K i values of amyloid radiotracers measured using synthetic tau and AD brain homogenates did not differ greatly [5]. e huge difference in the K D values measured using human brain homogenates and synthetic tau were much greater for [ 18 F] THK523 than for [ 18 F]THK5105 (Table 1). is might also indicate the binding preferences of [ 18 F]THK523 to certain tau-binding sites available on synthetic tau, that were fewer in numbers in human brain homogenates. erefore, it is important to determine the binding affinity of tau radiotracers to different subtypes of tau protein and other β-sheet structures such as Aβ and α-synuclein.
[ 18 F]THK5351 yielded higher clinically observed SUVR than [ 18 F]THK5117 in the same AD patients, with lower white matter binding [15]. [ 18 F]THK5351 was also reported to have a higher signal-to-noise ratio (SNR), and a lower non-specific binding in white matter than [ 18 (Table 3).
is difference may be attributed to the tau subtypes that [ 11 C]PBB3 is binding. [ 18 F] THK5351 and [ 18 F]flortaucipir was reported to bind to the same targets but with different affinities, while [ 11 C]PBB3 seems to bind to a different tau subtype [44]. If the tau subtype that [ 11 C]PBB3 binds to is of a lower concentration in subject, the clinical SUVR will become smaller. e difference between the clinically observed results and CUI ranking showed that the evaluation of the clinical usefulness of tau radiotracer based on binding to a single target could not reflect the actual in vivo binding in subjects. High tau selectivity and off-target binding affect the comparison of the in vivo binding of tau radiotracers, which are less prominent in amyloid radiotracers. Despite the differences in CUI rankings, the clinically applied tau radiotracers had CUI values above the recommended value especially for those with high SNR. us, the screening methodology can still provide confidence in the decision-making of moving candidate radiotracers for clinical studies.

Limitations of Screening Methodology.
Few measurements of tau concentration in postmortem human brains using ELISA have been reported, and these values are very different [17,29,45,46]. In addition, these reported tau concentrations were mostly measured using normal-aged control and AD brains, with very little data on the tau concentration in MCI. As such, the simulated SUVR distribution might not reflect the clinically observed MCI result. Moreover, the input function of the amyloid radiotracer [ 11 C]BF227 was used for simulations. us far, the input functions of only three clinically applied tau radiotracers of [ 11 C]PBB3 [10], [ 18 F]flortaucipir [36], and [ 18 F]MK6240 [16] have been reported. e arterial input functions of these radiotracers were similar in HC and AD, with a fast uptake and a fast washout, and the shape of the curves was similar to that of [ 11 C]BF227 as used in the simulation. Although the shape of the input function of these two radiotracers was similar to that of [ 11 C]BF227, the shape of the arterial input function might be different for other tau radiotracers. us, we evaluated the effect of the input function on the outcome using four different input functions with fast kinetics for HC and AD subjects injected with [ 11 C]BF227 or [ 18 F]FACT, with areas under the input function curves from 0 to 120 min of 536 (default), 649, 434, and 306 (kBq/mL) min. e % COV of the predicted SUVR was less than 7.0 for all conditions and radiotracers, while %COV of the CUI was less than 7.0 for all except the poor radiotracers, namely, [ 18  is showed that the results would not be changed significantly using input functions with similar kinetics. However, there were also issues with metabolites crossing the BBB (e.g., [ 11 C] PBB3), but the amyloid biomathematical screening methodology could not be used to predict the possibility of metabolites crossing this barrier.
Off-target binding was observed in some clinically applied tau radiotracers. [ 18 F]flortaucipir was reported to show specific binding in the midbrain, vessels, iron-associated regions (e.g., basal ganglia), substantia nigra, calcifications in the choroid plexus, and leptomeningeal melanin [13]. [ 11 C]PBB3 was reported to accumulate in the venous sinuses, basal ganglia, and thalamus, while its fluorinated compounds showed off-target binding in the choroid plexus [14,44]. [ 18 F]THK5351 was reported to bind to monoamine oxidase B (MAO-B), which is highly expressed throughout the brain, and thus, its tau binding data needs to be corrected for MAO-B binding [47]. [ 18 F]MK6240 was reported to have reduced off-target binding on the whole but showed offtarget binging in regions such as the retina, substantia nigra, ethmoid sinus, and dura matter [16]. Depending on the region of off-target binding, the effects may not limit PET quantification due to little or no anatomical overlap of the target regions of interest (ROIs) with off-target regions. Accurate PET quantification is also less affected if the radiotracer has high target selectivity or if the concentrations of the off-target binding sites are much lower compared to that of the target [48]. Off-target binding may be one of the contributing factors that led to the observed differences between simulation and the clinical data of tau PET radiotracers. e possibility of binding to off-targets is difficult to predict, and systematic screening is required to determine the binding of the candidate compound to a wide range of proteins. is will increase the time and cost of compound screening. e amyloid biomathematical screening methodology could not predict off-target binding, and the inclusion of multiple binding sites appeared to be required for tau radiotracers to correct for this issue.

Feasibility of Extending to the Screening of Tau
Radiotracers. To date, the comparison of multiple tau radiotracers has been performed via in vitro competition binding assays in human brain sections, using human AD brain homogenates [11,12] or by means of preclinical imaging [38]. ere is a lack of consideration of the possible in vivo kinetics of the radiotracers during development, which may lead to poor clinical performance [4][5][6]. e use of in silico data can support predictions of tracer kinetics and increases confidence in clinical translation, in addition to facilitating radiotracer comparisons. e weak SUVR correlation was obtained between the predicted and clinically observed SUVR results, mostly due to the small SUVR values for tau radiotracers with poorer tau selectivity. However, there are very few reported kinetic parameters to assess the limitations of the screening methodology. e TACs, SUVR distribution, and CUI rankings differed primarily for tau radiotracers with low selectivity to tau. is showed that the evaluation of the clinical usefulness of tau radiotracer based on binding to a single target could not fully reflect the actual in vivo binding in subjects since they also exhibited binding preferences to nontarget sites. us, it is not feasible to directly apply the amyloid biomathematical screening methodology to tau radiotracers due to the increased complexity of evaluating the binding of tau radiotracers, namely, target-binding, off-target binding, and non-specific binding. More work is required to improve the accuracy of predicting the clinical usefulness of tau radiotracers by including possible binding to other β-sheet structures or offtarget sites. However, the high CUI values generated for clinically applied tau radiotracers with high SNR showed that the screening methodology could be used to increase confidence in decision-making when choosing candidate radiotracers for further evaluation.

Conclusions
e predicted TACs, SUVR, and CUI ranking differed for some clinically applied tau radiotracers, especially those with lower selectivity for tau. is showed that the evaluation of the clinical usefulness of tau radiotracer based on binding to a single target could not reflect the actual in vivo tau binding in subjects due to more challenges in evaluating the in vivo binding of tau radiotracers, such as off-target binding and high tau selectivity, compared to amyloid radiotracers. e inclusion of possible binding to other β-sheet structures or off-target sites and the binding affinities to different target sites would improve the accuracy of the prediction. From our results, clinically applied tau radiotracers with higher SNR, such as [ 18 F]MK6240 and [ 18 F]THK5351, had higher CUI rankings. is supported the use of the screening methodology in radiotracer development by allowing comparison of candidate radiotracers with clinically-applied radiotracers based on SUVR, with respect to binding to a single target. Our results will hopefully provide some insights to guide the development of in silico models in supporting the development of tau radiotracers.
Data Availability e program (RSwCUI) used for TACs simulation and CUI evaluation can be download from http://www.rim.cyric. tohoku.ac.jp/software/CUI-Software. e predicted K1, k2, and BP ND values in HC and AD of 9 clinically applied taurelated radiotracers are provided in the interest of readers and are included within the supplementary information file.