Skip to contents

CD8 logistic regression score for paired TCRs

Source: R/cd8_scoring.R
make_cd8_score_table_column.Rd

Scores each TCR using a logistic regression model trained on sorted CD4/CD8 bulk TCR data. The alpha-chain and beta-chain models are fit independently and their scores are averaged (each weighted 0.5).

Usage

make_cd8_score_table_column(tcr_df, use_sigmoid = FALSE)

Arguments

tcr_df

A data.frame with columns va, ja, cdr3a, vb, jb, cdr3b. Gene names should include allele suffixes (e.g. "TRAV1-2*01").

use_sigmoid

Logical. If TRUE, apply the sigmoid function \(1 / (1 + exp(-x))\) to each chain's score before averaging. Default is FALSE.

Value

A numeric vector of CD8 logistic regression scores, one per row of tcr_df.

Details

The model encodes V-gene, J-gene, CDR3 length, and positional amino acid features using one-hot and k-hot encodings, then computes a dot product with pre-trained weights.

See also

match_tcrs_to_db

Examples

# \donttest{
tcr_df <- data.frame(
    va = "TRAV1-2*01", ja = "TRAJ33*01",
    cdr3a = "CAVMDSSYKLIF",
    vb = "TRBV6-4*01", jb = "TRBJ2-1*01",
    cdr3b = "CASSLAPGATNEKLFF",
    stringsAsFactors = FALSE
)
scores <- make_cd8_score_table_column(tcr_df)
# }