Application of machine learning and deep learning in processing biological data either in gene sequencing and expression or protein for solving some of the key challenges in life sciences has largely led to the application of data science to deal with data scientifically. Analysing clinical images can expide and improve accuracy in the pathological and radiological image processing and analysis.

Below is the list of some of the applications of AI models in scientific explorations…

1- Fiszmametal used NLP models for anti-infective therapies.

2- Miller et al used NLP models for automatic monitoring of adverse effects.

3- Castrol et al used NLP models for 14 cerebral aneurysms disease variables in clinical notes.

4- Afzal et al used NLP models for identification of peripheral artery disease related keywords in clinical notes.

5- Strokes costs 500 million USD in US and US $689 billion worldwide.

6- Viller et al used a genetic fuzzy finite state machine and PCA for predicting stroke.

7- Rehane et al used SVM models for predicting state on MRI data. Endophenotypes of motor disability after stroke were identified and classified.

8- Machine learning models were used to establish a direct relationship with intravenous thrombolysis with the prognosis and survival rate.

9- Benthley et al predicted symptomatic intracranial hemorrhage by CT scan.

10- Non synonymous variant was used to produce meta predictors to merge several predictions into one.

11- Combined annotation-dependent depletion approach (CADD) was used to predict deleteriousness of genetic data.

12- DWNN is not sufficient for clinical reporting but useful for guiding the interpretation of clinical genomic data by potential candidate variants for further consideration. Example primase AI, splicing defect detection genes – Splice AI, DeepSEA.

13- Prediction of DNase hypersensitive sites, transcription factor, binding sites, histone marks and genetic variations.

14- Biodion 1Hb sequence size improves performance.

15- Expecto uses gene expression level directly from DNA sequence information.

16- 1007 distinct terms defining abnormalities of face associated with 4526 diseases and 2142 genes.

17- DeepGestalt’s CNN based facial image precisely distinguishes between molecular diagnosis and clinical diagnosis.

18- PEDIA candidate pathogenic variant 105 different monogenetic disorders across 679 individuals.

19- Image driven phenotype identification and their probable source using CNN with Cox proportional hazard based outcomes. Histological features of brain tumors. Survival correlated with somatic mutation.

20- Feature subset 1,881 genes – F score algorithm – 93.3% accuracy on the cross validation set.

21- TE (Trophectoderm) – ICM which are two types Primitive endoderm and epiblast (EPI) cells.

22- ARGFX, CPHX1, LEUTX and DUX4 pioneering factors, overexpression experiments and transcriptional analysis.

23- In Mice, COX2 represses OCT4 expression in the outer cells, leading to TE-ICM linear segregation.

24- DPPA2 & DPPA4 regulate expression of DUX & LINE-1 in mouse embryonic stem cells.

25- QSPRED – FC Tools – Quorum sensing peptides.

26- DGE analysis edgeR, Limma and DeSEQ DPPA5, 2SCAN4 & SOX2.

The application of machine learning and deep learning is increasingly explored in many more areas within the life sciences industry either in pharma or provider or payer lines of business.

B3DS expertise in genomic and clinical data scientific exploration is helping industry unravel hidden biological mysteries at deep molecular level by analyzing tons of data at no cost and time leavering its highly scalable platform built on open Source stack and also leverages Databricks framework and open Source and custom built genomic data processing pipelines and algorithms.