TCRpower: quantifying the detection power of T-cell receptor sequencing with a novel computational pipeline calibrated by spike-in sequences

Abstract

T-cell receptor (TCR) sequencing has enabled the development of innovative diagnostic tests for cancers, autoimmune diseases and other applications. However, the rarity of many T-cell clonotypes presents a detection challenge, which may lead to misdiagnosis if diagnostically relevant TCRs remain undetected. To address this issue, we developed TCRpower, a novel computational pipeline for quantifying the statistical detection power of TCR sequencing methods. TCRpower calculates the probability of detecting a TCR sequence as a function of several key parameters: in-vivo TCR frequency, T-cell sample count, read sequencing depth and read cutoff. To calibrate TCRpower, we selected unique TCRs of 45 T-cell clones (TCCs) as spike-in TCRs. We sequenced the spike-in TCRs from TCCs, together with TCRs from peripheral blood, using a 5' RACE protocol. The 45 spike-in TCRs covered a wide range of sample frequencies, ranging from 5 per 100 to 1 per 1 million. The resulting spike-in TCR read counts and ground truth frequencies allowed us to calibrate TCRpower. In our TCR sequencing data, we observed a consistent linear relationship between sample and sequencing read frequencies. We were also able to reliably detect spike-in TCRs with frequencies as low as one per million. By implementing an optimized read cutoff, we eliminated most of the falsely detected sequences in our data (TCR α-chain 99.0% and TCR β-chain 92.4%), thereby improving diagnostic specificity. TCRpower is publicly available and can be used to optimize future TCR sequencing experiments, and thereby enable reliable detection of disease-relevant TCRs for diagnostic applications.

Keywords: T-cell receptor; TCRpower and adaptive immune receptor repertoire sequencing; bulk T-cell receptor sequencing; computational model; spike-in standards.

Figure 1

Study design. Our study presents a detection power calculator based on a computational model of TCR RNA read count in bulk sequencing data to enable efficient TCR sampling and RNA sequencing. Our model has two components [1] modeling the number of target TCR in a peripheral blood sample and [2] modeling the number of target TCR RNA sequencing reads based on the number of sampled target TCR. To calibrate our model, we mixed RNA from T cells with known (spike-in) concentration, together with RNA from CD4 effector memory T cells with unknown TCRs, and sequenced TCR from these mixtures using a 5′ RACE protocol. To investigate how library preparation choices affect detection power, we created three sequencing sets with different library preparation approaches. As controls, we performed TCR sequencing on the spike-in RNA mix only (Control spike-in TCC mix) and RNA of the effector memory CD4 T cells only (Control CD4 TEM). Created with Biorender.com.

Figure 2

Accuracy and variability in TCR frequency measurement. (A) Ground truth versus measured TCR frequency of the spike-in TCR for experimental Sets 1–3 for all 6 replicates, showing consistent linear relationships. (B) TCR frequency dispersion index (std divided by mean) across 6 replicates of each TCR. The lower frequency TCRs have higher dispersion (R²) than the higher frequency TCRs. Some low-frequency TCRs are undetected (marked by X) either for a certain replicate (panel A) or across all six replicates (panel B).

Figure 3

Negative binomial modeling of the spike-in TCR read counts and detection limit estimation. (A) TRA (red) and TRB (blue) read count versus spike-in TCR frequency under three experimental conditions (Sets 1–3). The dots represent measured read counts, whereas the green line and areas are the respective mean and 95% prediction interval of negative binomial models with read efficiency parameter r_e and mean-variance relationship parameters η, λ, fitted by maximum likelihood. T_read = the total TRA or TRB read count for the set. (B) Estimated detection probability (read count > 0) as a function of TCR frequency, along with the minimal fraction (dashed line) that can be detected with at least 95% probability. Note that for TRA, Set 3 stands out with the lowest η value (i.e. variance) and 95% detection probability.

Figure 4

Falsely detected sequences in the Control Spike-in TCC set. (A) Count distributions of TRA and TRB reads that did not match the TCRs of the spike-in TCCs, but were nevertheless detected in the Control Spike-in TCC set. All of the falsely detected sequences were either present in other sets in the library (purple), or only exclusively found in the Control Spike-in TCC set (green). The dotted lines represent the read cutoff [18] (B) Read count total over all CD4 TEM containing sets versus read counts in the Control Spike-in TCC Mix for the falsely detected sequences with read count >18 in the Control Spike-in TCC Mix. Note the linear trend characteristic of the index-hopping phenomenon.

Figure 5

Detection power estimation. Example output from our power calculator TCRpower, showing the probability of detecting a TCR with frequency 10^–4 and read count threshold 18, as a function of the number of sampled TCR and sequencing reads.