Counting colonies on agar plates is a widely used method in microbiology.OpenCFU is a free software that should facilitate (and render more reproducible) the enumeration of colony forming unit (CFU).You can simply run the program on your computer and input pictures of plated bacterial colonies (or other cells).
Secure banking anytime, anywhere with 24/7 access to your account with our online and mobile banking. Check balances, view transactions, get instant alerts, pay bills, transfer funds or make mobile deposits through our convenient app.
Colony online free
One issue that appears to limit iPSC quality or biological consistency in further applications is colony determination, which leads to the isolation and purification of iPSC colonies during the cellular reprogramming process. The current solution relies on a judgement call from well-trained cell culture experts, though the training is time consuming and has a high cost. In contrast and inevitably, the natural instability of human recognition will could allow for misjudgement, which can cause batch-to-batch variation. The quality control of iPSC colony determination is extremely important for downstream expansion and for maintaining a homogeneous culture of undifferentiated cells. Inconsistency in colony determination and selection will cause insufficient downstream differentiation into functional cells. Therefore, an automated quantitative method with consistent colony maturation determination will be a great for assisting biologists during the iPSC formation process.
The traditional approach for identifying and verifying iPSC colonies is to use immunofluorescence staining or a reporter system to detect pluripotent markers, which could couple well with automated fluorescence microscopy provides and provide a dynamic and effective method4,5. However, this kind of xeno probe-based labelling can only be applied during a late stage in reprogramming; this cannot be delayed further due to potential safety issues for downstream clinical applications6,7. Furthermore, during cellular reprogramming studies, it has been observed that along with positive iPSC colonies, negative colonies contain various morphology subtypes, especially with respect to cellular polarity. The cellular polarity is strongly linked to the specificity of gene expression, cell cycle, and other cellular regulation and may explain the different reprogramming mechanisms8. For example, the cell polarity changes between the mesenchymal-to-epithelial transition (MET) or EMT, which is linked with induced pluripotent stem cell (iPSC) colony formation or tumour genesis9,10,11. Hence, the reprogramming process has a strong link with morphology changes; the classification of the morphology can be used as a read-out in a quantitative way to identify iPSC colonies and to monitor the reprogramming process. This provides richer information than the binary fluorescence images and opens the door for a label-free, non-invasive approach.
During the reprogramming process, the cellular polarization reshaping leads to morphology changes, which indicate iPSC colony formation. A novel computer vision assisted method was developed to process the microscopic images. The sliding window approach was performed to scan the potential colony areas, followed by the detection, which generates a binary image. Post-processing including discarding the fragmented areas was performed using the binary images as the template. Therefore, the areas of interest referring to the iPSC colonies can be determined based on size and location (Fig. 1b). Examples of cropped positive and negative colony texture mosaic windows from the algorithm training is shown in Fig. 1a. The algorithm detects the positive areas in a whole well from a 6-well cell culture plate.
Human iPSC regeneration opens the door as a personalized cell source for cell replacement therapy in regenerative medicine. This algorithm has extended the non-invasive and label-free computer assisted method for human urinal cell reprogramming3. Different from the OG mice system, since urinal cells OCT4-GFP is not applicable in this case, we built the training dataset using manual cropping processed by well-trained cell biologists (Fig. 2a). As shown in the previous section, our iPSC detection framework consists of a mosaic sliding window-based classification process. If a small patch is located in a positive iPSC colony texture, the area was then masked out by the overlay; otherwise, it is a negative patch. The training session was run with numerous manually cropped samples, followed by the annotation of the positive colonies; this indicated the locations of regions after post-processing for an entire well in a 6-well plate (Fig. 2b).
This computer-guided selection was implemented after combining the generated binary image with the bright-field image, and the colony-picking decisions were made only based on the computer vision results. Subsequently, to verify the computer-guided iPSC colony detection, manual picking was performed, followed by standard iPS cell characterization (Fig. 2c,d). The expanded colonies that passed the characterization, including karyotyping by G-band analysis (data not shown), were further characterized for pluripotency. Here, we show iPSCs generated from sample c4p2. For example, fluorescent immunostaining was initially performed to demonstrate the expression of pluripotent surface markers, such as TRA-1-60, TRA-1-81 and SSEA4 (Fig. 2c). Then, the endogenous pluripotent genes, such as OCT4, SOX2 and NANOG, were fully activated, and were comparable to human embryonic stem line H1 cells (Fig. 2d). Finally, using genomic PCR, it was confirmed that the UC-iPSCs expanded at passage 21 and no longer harboured the exogenous reprogramming factors from the original episomal plasmid (Fig. 2e). Hence, this computer vision approach worked for both human iPSC detection as well as in the mouse model.
First, an individual colony was back traced as described early on. Combined with the detection method, the earliest feature for cellular polarity changes, the formation of the iPSC texture can be recognized at day 7 after reprogramming induction. An example of the detected time-lapse data is shown in Fig. 3a. Subsequently, the individual iPSC texture features were detected using a segmented boundary, and each feature was registered and tagged digitally (Fig. 3b). Each iPSC texture feature was monitored individually during the entire reprogramming process. We can easily exclude the over-grown and under-grown colonies using the average growth rate. This gives a quantitative measurement for colony formation, which is shown in the growth phases, and each curve line represents a qualified iPSC clone (Fig. 3c). Moreover, a Hidden Markov Model (HMM) with four quantitative features was applied to analyse this process. This means that under the comparable reprogramming conditions with the same cells types, the model classified four different phases in the growth curve to describe the characterization of the iPSC clone forming mathematically. Finally, this model gives a posterior probability for each phase; for example, between days 12 and 17, the probability score of maturation phase increased from 0 to 1. The model provides the correlation between the iPSC harvest time and the probability score. The image data for best picking decision was selected manually and fed into the model to calculate the selection threshold. The probability score of 0.3 was the output to describe the picking threshold in the urinal cell reprogramming system. For example, in this data set, the algorithm gives the closest score to 0.3 on day 14, which means that colony picking could be triggered. Therefore, the optimal picking time window can be predicted. If the probability factor reaches 1, this means the cells are overgrown, which implies a risk of random differentiation.
In separate experiments, iPSC colony detection was synchronized using a clone-by-clone approach, and the colony selection was triggered as described above. Subsequently, these colonies were verified using a sample enriched for high-throughput RNA-Seq gene expression analysis. The pluripotency- and germ layer-specific marker genes were plotted for comparison.
Deep learning has recently attracted attention in the machine learning field. The motivation is to use big data to directly train multi-layer neural networks with different deep structures and combine these networks with feature extraction and classification. Convolution Neural Network (CNN) is a structure introduced by LeCun27 that is widely used in computer vision area. Its most successful application is dealing with image data in classification, segmentation detection and retrieval task. For example, CNN is used as the current baseline approach in breast cancer classification and diagnosis28. Chen et al.29 present a multi-feature, label-free cell classification system using deep learning techniques. The beauty of the deep learning technique is the end-to-end learning procedure that combines feature extraction with the classifier. The performance will be benefited by expanding training data set.
Here, we present a machine learning-based system for the automated detection and prediction of iPSC formation using a cellular reprogramming process. This model works in both mice and human reprogramming systems, and the HMM-based model was used for phase probability estimation to trigger iPSC colony selection. Our approach, which uses a Deep Convolution Neural Network (DCNN) end-to-end learning framework, can avoid the non-optimal manual design of extractors and classifiers when faced with complicated cell textures and morphology changes, which provides optimized performance and convenience.
Finally, the whole reprogramming process is serum-free, feeder-free and uses episomal based induction, this computer vision guided label free and non-invasive approach was fully verified by standard biological approaches, as well as RNA sequencing. We expect these combined approach will become an everyday technique for cell biology studies in a quantitative way. It should not limited in cellular reprogramming works. This system can be developed further includes to study downstream cell differentiation, and cell line development to identify appropriate cells in a fully traceable and quantitative way, or even guide an automated robotic in application of regenerative medicine. 2ff7e9595c
コメント