Antibody Stuff

Big Data Analysis of Antibodies Leading to Production of Monoclonal Antibodies (or polyclonal, those are useful too)

GenScript does work with antibodies

We can do antibodies in most "wet" labs.

Author: 	Guang Lan Zhang
Article title: 	Big Data Analytics in Immunology: A Knowledge-Based Approach
URL: https://www.hindawi.com/journals/bmri/2014/437987/

Article title: 	Monoclonal Antibody “Gold Rush”: Ingenta Connect
Website title: 	Ingentaconnect.com
URL: 	https://www.ingentaconnect.com/content/ben/cmc/2007/00000014/00000018/art00008

Author: Anish K. Simhal
Article title: 	A Computational Synaptic Antibody Characterization Tool for Array Tomography
URL: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6057115/

Author: Elizabeth Rossin
Article title: 	A framework for analytical characterization of monoclonal antibodies based on reactivity profiles in different tissues
URL: https://academic.oup.com/bioinformatics/article/27/19/2746/232045

Author:	Christoph J. Blohmke
Article title: 	The use of systems biology and immunological big data to guide vaccine development
URL: https://genomemedicine.biomedcentral.com/articles/10.1186/s13073-015-0236-1

Author 	Ngoc Hieu Tran
Article title: 	Complete De Novo Assembly of Monoclonal Antibody Sequences
URL: https://www.nature.com/articles/srep31730

Website title: 	Machine learning for classification and quantification of monoclonal antibody preparations for cancer therapy
URL: 	https://arxiv.org/pdf/1705.07099.pdf

Author 	Rapid Inc
Article title: 	A Single Platform to Sequence All Monoclonal Antibody Proteins
Website title: 	News-Medical.net
URL: 	https://www.news-medical.net/news/20181003/A-Single-Platform-to-Sequence-All-Monoclonal-Antibody-Proteins.aspx


Introduction to Antigens

Introduction to Big Data

Hybridomas still necessary

What techniques have been proposed to address the antibody protein sequencing problem?

Over the years, several papers have been published to address the antibody protein sequencing problem. From the manual sequencing and assembly approach published 25 years ago, to the homology database assisted sequencing algorithms that can achieve over 90% accuracy, to the self-claimed automated full-length sequencing software released in recent years. But none of them have been widely adopted in the real world.

 Agile methodology

Rapid Novor developments

An Expectation–Maximization (EM) algorithm is used for fitting every pattern with a smooth and mathematically rigorous profile that specifies all key features precisely, with the help of a probability density function. In the second step, for each tissue, mAbprofiler performs curve clustering of the fitted mAb profiles with a novel of skew t mixture of non-linear regression model that is robust against intersample variation. We used an effective criterion, the Jump Statistic, for model selection with the optimal number of clusters (or mAb classes).

Jump function algorithms

Besides internal validation, we also compared the performance of mAbprofiler with other established methods. We began with hierarchical clustering, which is the most commonly used approach for mAb classification (Bernard and Boumsell, 1984). When we used hierarchical clustering on our mAb profiles, then the method clearly failed to capture the complex class structure and detected few clusters. Based on Average Silhouette Width (ASW), a common measure for determining the quality of hierarchical clustering, we noted that the optimal number of mAb classes according to hierarchical clustering of our data was typically restricted to four or even fewer for all tissues other than thymocytes. Moreover, little difference among the ASW scores for different number of clusters indicated that the hierarchical clusters had low separation (Supplementary Figure S5).

Thereafter, we adopted the established protocol of Pratt et al. (2009) in which mAb histograms were first smoothed with SiZer, and then hierarchical clustering was performed with those smoothed profiles. We show the results of that approach on our data using SiZer plots for the different tissues in Supplementary Figure S6a–f. As depicted with the dendrograms, while the larger clustering structures were detected with smoothing, the finer structures were often ignored, thus resulting in highly heterogeneous classes. This can be seen clearly in the largest clusters in spleen, lung lavage and bone marrow.




https://ramp.studio/ -

CV bagging

The overall winner of the contest was the Mines RAMP with a classification error of 0.7%
(only three low-concentration test samples were missed!)  and a concentration MARE of 5.8%.

from Raman spectroscopy