Methods to incorporate machine learning analytics for optimizing protein purity, potency and quality in an on-demand production system for point-of-care delivery
The present invention relates to cell free protein manufacturing, and more particularly, for integrating machine learning into a portable cell-free bioprocessing system for producing proteins with increased and consistent purity, potency and quality wherein such proteins are prepared on-demand and for point-of-care delivery.
1 . A method of analyzing and estimating purity of a protein produced in a portable, cell-free bioprocessing system, wherein the bioprocessing system comprises a protein expression module and a protein purification module, wherein each module is associated with on-board analytics, the method comprising:
producing the protein in the protein expression module;
purifying the produced protein using a two-step purification process in the protein purification module, wherein the protein purification module comprises a UV sensor for each purification process and a multiplicity of programmable syringe pumps with integrated pressure sensors positioned at a back-end of each programmable syringe pump;
obtaining at least pressure and UV sensor data from the protein purification module during purification of the produced protein;
transmitting the data to a computer aided classification system;
extracting features from the data with the computer aided classification system for classifying the produced protein and process conditions, wherein extracted features characterize the produced protein and such sample characterization is compared to characterization of extracted features from previously prepared proteins to provide classified features of the produced protein; and
applying an unsupervised clustering process to the classified features to provide a plurality of output clusters to provide an estimation of the purity of the produced protein using in-line analysis.
2 . The method of claim 1 , wherein an AI machine learning module collects and stores in-line real-time testing data of purified protein from the protein purification module and further provides information on product quality and potency for each batch of the produced proteins relative to previously produced proteins.
3 . The method of claim 1 , wherein the protein expression module comprises at least one dialysis cassette or reactor for inclusion of cell lysate, a reaction mixture, and DNA or mRNA for production of the protein.
4 . The method of claim 1 , wherein the pressure sensors and UV sensors monitor each step of the purification process.
5 . The method of claim 4 , wherein the programmable syringe pumps dispense buffers.
6 . The method of claim 3 , wherein the cell lysate is from CHO cells or E. coli cells.
7 . The method of claim 3 , wherein the reaction mixture comprises at least one of amino acids, nucleotides, co-factors, enzymes, ribosomes, tRNA, polymerases, and transcriptional factors.
8 . The method of claim 1 , wherein the protein purification module comprises a metal ion affinity chromatography column for initial purification and an ion-exchange chromatography column for a polishing step.
9 . The method of claim 4 , wherein each step of the purification process comprises an inline flow cell equipped with an inline UV sensor comprising light sources and detectors for measuring UV absorbance at about 280 nm to monitor the purification process.
10 . The method of claim 1 , further comprising measuring dissolved oxygen, pH, absorbance, pressure and temperature during the production process.
11 . The method of claim 2 , wherein the AI machine learning module uses a blind source separation (BSS) algorithm.
12 . The method of claim 11 , wherein the (BSS) algorithm uses independent-component analysis (ICA) that extracts independent source signals when the source signals are active simultaneously and is a BSS algorithm depending on using Artificial Neural Networks.
13 . The method of claim 4 , wherein each of the two purification processes comprise at least one real-time test selected from the group consisting of absorbance, circular dichroism, fluorescence measurements, and lifetime measurements.
14 . The method of claim 2 , wherein the AI machine learning module is a cloud-based server or a physical server connected to the bioprocessing system.
15 . The method of claim 1 , wherein the bioprocessing system further comprises a microfluidic mixer positioned between the protein expression module and the protein purification module to mix expressed protein with buffer.
16 . The method of claim 1 , wherein UV and pressure sensor profiles are used to detect production deviations and reject a batch of the produced protein.
17 . The method of claim 1 , wherein the bioprocessing system further comprises a processor for controlling and/or monitoring at least the protein expression module and/or the protein purification module.
18 . The method of claim 17 , wherein the processor can control a length of time in the protein expression module and/or purification module to affect potency and/or activity of the produced protein.
19 . The method of claim 1 , wherein the bioprocessing system comprises a single-use bioprocessing train comprising at least a bioreactor, tubing, at least one microfluidic mixer, a capture purification column, and a polishing purification column.
20 . The method of claim 1 , wherein the purification process comprises affinity chromatography as a first step and ion-exchange chromatography as a second step, wherein the protein purification module comprises a built-in flow cell equipped with the respective UV sensor, and wherein the pressure sensors provide continuous system pressure monitoring.