Every possible sequence, accounted for
Ask how likely a biological sequence is, how deeply to sequence, how many people share a sequence, or how two repertoires, cell types or conditions differ. BEAM learns from your sequences alone, with their abundances if you have them, and answers by summing over every combinatorially possible sequence, without listing or sampling a single one.
What BEAM answers
Section titled “What BEAM answers”How likely is this sequence?The probability of any sequence, normalised over all possible sequences, including ones never seen before: for example, of a CDR3β under a reference model of human repertoires, and which residues make it likely or unlikely.
How deeply should I sequence?How many distinct sequences to expect at a given depth, how often sequences repeat, and how many two samples will have in common.
How many people share this sequence?The expected number of repertoires in a cohort that contain a sequence, the full distribution around it, and how many of a cohort's sequences are private.
How different are two repertoires?A distance that separates differences in sequence length from differences in local sequence patterns, and names the patterns.
How do cell types or conditions differ?Group models for CD4 and CD8, naive and memory, tissues or cohorts: compare groups, cluster repertoires, and see which group a sample resembles.
How diverse is my sample, really?The population's diversity, how much of it a sample of your depth shows, and the chance that the next cell is new.
See every question BEAM answers →
Install
Section titled “Install”pip install beam-modelBEAM runs on Linux, macOS and Windows with Python 3.11 to 3.15. It comes with a reference model built from 38,538 human repertoires, ready to load in one line.
T. Konstantinovsky and G. Yaari. A computable probability landscape for T cell receptor repertoire sampling and comparison. (Submitted.)