Statistical classifiers for diagnosing disease from immune repertoires

jostmey · September 18, 2017, 6:39pm

I want to share our preliminary work on developing statistical classifiers for repertoires. The main idea in our paper is to score snippets (6-mers) of CDR3 sequence by their biochemical features with a “detector” function and to aggregate the scores into a single value that can represent a diagnosis. We believe this is an important step toward utilizing the information contained in each individual sequence instead of relying on summary statistics of repertoires (i.e. diversity scores).

ncbi.nlm.nih.gov

Statistical classifiers for diagnosing disease from immune repertoires: a case study using multiple sclerosis.

J Ostmeyer, S Christley, WH Rounds, I Toby, BM Greenberg, NL Monson and LG Cowell, BMC bioinformatics, Sep 2017 07

Deep sequencing of lymphocyte receptor repertoires has made it possible to comprehensively profile the clonal composition of lymphocyte populations. This opens the door for novel approaches to diagnose and prognosticate diseases with a driving immune component by identifying repertoire sequence patterns associated with clinical phenotypes. Indeed, recent studies support the feasibility of this, demonstrating an association between repertoire-level summary statistics (e.g., diversity) and patient outcomes for several diseases. In our own prior work, we have shown that six codons in VH4-containing genes in B cells from the cerebrospinal fluid of patients with relapsing remitting multiple sclerosis (RRMS) have higher replacement mutation frequencies than observed in healthy controls or patients with other neurological diseases. However, prior methods to date have been limited to focusing on repertoire-level summary statistics, ignoring the vast amounts of information in the millions of individual immune receptors comprising a repertoire. We have developed a novel method that addresses this limitation by using innovative approaches for accommodating the extraordinary sequence diversity of immune receptors and widely used machine learning approaches. We applied our method to RRMS, an autoimmune disease that is notoriously difficult to diagnose.We use the biochemical features encoded by the complementarity determining region 3 of each B cell receptor heavy chain in every patient repertoire as input to a detector function, which is fit to give the correct diagnosis for each patient using maximum likelihood optimization methods. The resulting statistical classifier assigns patients to one of two diagnosis categories, RRMS or other neurological disease, with 87% accuracy by leave-one-out cross-validation on training data (N = 23) and 72% accuracy on unused data from a separate study (N = 102).Our method is the first to apply statistical learning to immune repertoires to aid disease diagnosis, learning repertoire-level labels from the set of individual immune repertoire sequences. This method produced a repertoire-based statistical classifier for diagnosing RRMS that provides a high degree of diagnostic capability, rivaling the accuracy of diagnosis by a clinical expert. Additionally, this method points to a diagnostic biochemical motif in the antibodies of RRMS patients, which may offer insight into the disease process.

wsdewitt · September 18, 2017, 9:49pm

Neat paper!

From your Abstract:

…prior methods to date have been limited to focusing on repertoire-level summary statistics, ignoring the vast amounts of information in the millions of individual immune receptors comprising a repertoire. We have developed a novel method that addresses this limitation by using innovative approaches for accommodating the extraordinary sequence diversity of immune receptors and widely used machine learning approaches.

And from the Conclusions:

Our method is the first to apply statistical learning to immune repertoires to aid disease diagnosis, learning repertoire-level labels from the set of individual immune repertoire sequences.

You might be interested in these papers:

Thomas et al. Bioinformatics, Volume 30, Issue 22, 15 November 2014, where Atchley vector encoding of TCR sequence kmers is used to classify immunization exposure
Emerson et al. Nature Genetics 49, 659–665 (2017), where a viral infection classifier built from ~100 million TCR sequence features from ~700 subjects performs with AUROC > 0.9 in cross validation and in a separate validation cohort.
Dash et al. Nature 547, 89–93 (06 July 2017) and Glanville et al. Nature 547, 94–98 (06 July 2017), each of which presents methods to classify epitope specificity of TCR repertoires.

jostmey · September 19, 2017, 3:25am

Thanks.

Perhaps we should have phrased that sentence (and maybe another one like it) differently. The last two papers came out after we had submitted for peer-review.

I think our approach is distinct in that it really highlights using sequence level features, not features that summarize a cluster or that summarize a repertoire.